Commit | Line | Data |
---|---|---|
0a8165d7 | 1 | /* |
eb47b800 JK |
2 | * fs/f2fs/data.c |
3 | * | |
4 | * Copyright (c) 2012 Samsung Electronics Co., Ltd. | |
5 | * http://www.samsung.com/ | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License version 2 as | |
9 | * published by the Free Software Foundation. | |
10 | */ | |
11 | #include <linux/fs.h> | |
12 | #include <linux/f2fs_fs.h> | |
13 | #include <linux/buffer_head.h> | |
14 | #include <linux/mpage.h> | |
15 | #include <linux/writeback.h> | |
16 | #include <linux/backing-dev.h> | |
8f46dcae | 17 | #include <linux/pagevec.h> |
eb47b800 JK |
18 | #include <linux/blkdev.h> |
19 | #include <linux/bio.h> | |
690e4a3e | 20 | #include <linux/prefetch.h> |
e2e40f2c | 21 | #include <linux/uio.h> |
fe76b796 JK |
22 | #include <linux/mm.h> |
23 | #include <linux/memcontrol.h> | |
f1e88660 | 24 | #include <linux/cleancache.h> |
174cd4b1 | 25 | #include <linux/sched/signal.h> |
eb47b800 JK |
26 | |
27 | #include "f2fs.h" | |
28 | #include "node.h" | |
29 | #include "segment.h" | |
db9f7c1a | 30 | #include "trace.h" |
848753aa | 31 | #include <trace/events/f2fs.h> |
eb47b800 | 32 | |
36951b38 CY |
33 | static bool __is_cp_guaranteed(struct page *page) |
34 | { | |
35 | struct address_space *mapping = page->mapping; | |
36 | struct inode *inode; | |
37 | struct f2fs_sb_info *sbi; | |
38 | ||
39 | if (!mapping) | |
40 | return false; | |
41 | ||
42 | inode = mapping->host; | |
43 | sbi = F2FS_I_SB(inode); | |
44 | ||
45 | if (inode->i_ino == F2FS_META_INO(sbi) || | |
46 | inode->i_ino == F2FS_NODE_INO(sbi) || | |
47 | S_ISDIR(inode->i_mode) || | |
48 | is_cold_data(page)) | |
49 | return true; | |
50 | return false; | |
51 | } | |
52 | ||
4246a0b6 | 53 | static void f2fs_read_end_io(struct bio *bio) |
93dfe2ac | 54 | { |
f568849e LT |
55 | struct bio_vec *bvec; |
56 | int i; | |
93dfe2ac | 57 | |
8b038c70 | 58 | #ifdef CONFIG_F2FS_FAULT_INJECTION |
55523519 CY |
59 | if (time_to_inject(F2FS_P_SB(bio->bi_io_vec->bv_page), FAULT_IO)) { |
60 | f2fs_show_injection_info(FAULT_IO); | |
8b038c70 | 61 | bio->bi_error = -EIO; |
55523519 | 62 | } |
8b038c70 CY |
63 | #endif |
64 | ||
4375a336 | 65 | if (f2fs_bio_encrypted(bio)) { |
4246a0b6 | 66 | if (bio->bi_error) { |
0b81d077 | 67 | fscrypt_release_ctx(bio->bi_private); |
4375a336 | 68 | } else { |
0b81d077 | 69 | fscrypt_decrypt_bio_pages(bio->bi_private, bio); |
4375a336 JK |
70 | return; |
71 | } | |
72 | } | |
73 | ||
12377024 CY |
74 | bio_for_each_segment_all(bvec, bio, i) { |
75 | struct page *page = bvec->bv_page; | |
f1e88660 | 76 | |
4246a0b6 | 77 | if (!bio->bi_error) { |
237c0790 JK |
78 | if (!PageUptodate(page)) |
79 | SetPageUptodate(page); | |
f1e88660 JK |
80 | } else { |
81 | ClearPageUptodate(page); | |
82 | SetPageError(page); | |
83 | } | |
84 | unlock_page(page); | |
85 | } | |
f1e88660 JK |
86 | bio_put(bio); |
87 | } | |
88 | ||
4246a0b6 | 89 | static void f2fs_write_end_io(struct bio *bio) |
93dfe2ac | 90 | { |
1b1f559f | 91 | struct f2fs_sb_info *sbi = bio->bi_private; |
f568849e LT |
92 | struct bio_vec *bvec; |
93 | int i; | |
93dfe2ac | 94 | |
f568849e | 95 | bio_for_each_segment_all(bvec, bio, i) { |
93dfe2ac | 96 | struct page *page = bvec->bv_page; |
36951b38 | 97 | enum count_type type = WB_DATA_TYPE(page); |
93dfe2ac | 98 | |
0a595eba JK |
99 | if (IS_DUMMY_WRITTEN_PAGE(page)) { |
100 | set_page_private(page, (unsigned long)NULL); | |
101 | ClearPagePrivate(page); | |
102 | unlock_page(page); | |
103 | mempool_free(page, sbi->write_io_dummy); | |
104 | ||
105 | if (unlikely(bio->bi_error)) | |
106 | f2fs_stop_checkpoint(sbi, true); | |
107 | continue; | |
108 | } | |
109 | ||
0b81d077 | 110 | fscrypt_pullback_bio_page(&page, true); |
4375a336 | 111 | |
4246a0b6 | 112 | if (unlikely(bio->bi_error)) { |
5114a97a | 113 | mapping_set_error(page->mapping, -EIO); |
38f91ca8 | 114 | f2fs_stop_checkpoint(sbi, true); |
93dfe2ac | 115 | } |
36951b38 CY |
116 | dec_page_count(sbi, type); |
117 | clear_cold_data(page); | |
93dfe2ac | 118 | end_page_writeback(page); |
f568849e | 119 | } |
36951b38 | 120 | if (!get_pages(sbi, F2FS_WB_CP_DATA) && |
f5730184 | 121 | wq_has_sleeper(&sbi->cp_wait)) |
93dfe2ac JK |
122 | wake_up(&sbi->cp_wait); |
123 | ||
124 | bio_put(bio); | |
125 | } | |
126 | ||
3c62be17 JK |
127 | /* |
128 | * Return true, if pre_bio's bdev is same as its target device. | |
129 | */ | |
130 | struct block_device *f2fs_target_device(struct f2fs_sb_info *sbi, | |
131 | block_t blk_addr, struct bio *bio) | |
132 | { | |
133 | struct block_device *bdev = sbi->sb->s_bdev; | |
134 | int i; | |
135 | ||
136 | for (i = 0; i < sbi->s_ndevs; i++) { | |
137 | if (FDEV(i).start_blk <= blk_addr && | |
138 | FDEV(i).end_blk >= blk_addr) { | |
139 | blk_addr -= FDEV(i).start_blk; | |
140 | bdev = FDEV(i).bdev; | |
141 | break; | |
142 | } | |
143 | } | |
144 | if (bio) { | |
145 | bio->bi_bdev = bdev; | |
146 | bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(blk_addr); | |
147 | } | |
148 | return bdev; | |
149 | } | |
150 | ||
151 | int f2fs_target_device_index(struct f2fs_sb_info *sbi, block_t blkaddr) | |
152 | { | |
153 | int i; | |
154 | ||
155 | for (i = 0; i < sbi->s_ndevs; i++) | |
156 | if (FDEV(i).start_blk <= blkaddr && FDEV(i).end_blk >= blkaddr) | |
157 | return i; | |
158 | return 0; | |
159 | } | |
160 | ||
161 | static bool __same_bdev(struct f2fs_sb_info *sbi, | |
162 | block_t blk_addr, struct bio *bio) | |
163 | { | |
164 | return f2fs_target_device(sbi, blk_addr, NULL) == bio->bi_bdev; | |
165 | } | |
166 | ||
940a6d34 GZ |
167 | /* |
168 | * Low-level block read/write IO operations. | |
169 | */ | |
170 | static struct bio *__bio_alloc(struct f2fs_sb_info *sbi, block_t blk_addr, | |
171 | int npages, bool is_read) | |
172 | { | |
173 | struct bio *bio; | |
174 | ||
740432f8 | 175 | bio = f2fs_bio_alloc(npages); |
940a6d34 | 176 | |
3c62be17 | 177 | f2fs_target_device(sbi, blk_addr, bio); |
940a6d34 | 178 | bio->bi_end_io = is_read ? f2fs_read_end_io : f2fs_write_end_io; |
12377024 | 179 | bio->bi_private = is_read ? NULL : sbi; |
940a6d34 GZ |
180 | |
181 | return bio; | |
182 | } | |
183 | ||
4fc29c1a LT |
184 | static inline void __submit_bio(struct f2fs_sb_info *sbi, |
185 | struct bio *bio, enum page_type type) | |
f5730184 | 186 | { |
4fc29c1a | 187 | if (!is_read_io(bio_op(bio))) { |
0a595eba JK |
188 | unsigned int start; |
189 | ||
0bfd7a09 | 190 | if (f2fs_sb_mounted_blkzoned(sbi->sb) && |
52763a4b | 191 | current->plug && (type == DATA || type == NODE)) |
19a5f5e2 | 192 | blk_finish_plug(current->plug); |
0a595eba JK |
193 | |
194 | if (type != DATA && type != NODE) | |
195 | goto submit_io; | |
196 | ||
197 | start = bio->bi_iter.bi_size >> F2FS_BLKSIZE_BITS; | |
198 | start %= F2FS_IO_SIZE(sbi); | |
199 | ||
200 | if (start == 0) | |
201 | goto submit_io; | |
202 | ||
203 | /* fill dummy pages */ | |
204 | for (; start < F2FS_IO_SIZE(sbi); start++) { | |
205 | struct page *page = | |
206 | mempool_alloc(sbi->write_io_dummy, | |
207 | GFP_NOIO | __GFP_ZERO | __GFP_NOFAIL); | |
208 | f2fs_bug_on(sbi, !page); | |
209 | ||
210 | SetPagePrivate(page); | |
211 | set_page_private(page, (unsigned long)DUMMY_WRITTEN_PAGE); | |
212 | lock_page(page); | |
213 | if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) | |
214 | f2fs_bug_on(sbi, 1); | |
215 | } | |
216 | /* | |
217 | * In the NODE case, we lose next block address chain. So, we | |
218 | * need to do checkpoint in f2fs_sync_file. | |
219 | */ | |
220 | if (type == NODE) | |
221 | set_sbi_flag(sbi, SBI_NEED_CP); | |
19a5f5e2 | 222 | } |
0a595eba | 223 | submit_io: |
554b5125 JK |
224 | if (is_read_io(bio_op(bio))) |
225 | trace_f2fs_submit_read_bio(sbi->sb, type, bio); | |
226 | else | |
227 | trace_f2fs_submit_write_bio(sbi->sb, type, bio); | |
4e49ea4a | 228 | submit_bio(bio); |
f5730184 JK |
229 | } |
230 | ||
458e6197 | 231 | static void __submit_merged_bio(struct f2fs_bio_info *io) |
93dfe2ac | 232 | { |
458e6197 | 233 | struct f2fs_io_info *fio = &io->fio; |
93dfe2ac JK |
234 | |
235 | if (!io->bio) | |
236 | return; | |
237 | ||
554b5125 JK |
238 | bio_set_op_attrs(io->bio, fio->op, fio->op_flags); |
239 | ||
04d328de | 240 | if (is_read_io(fio->op)) |
554b5125 | 241 | trace_f2fs_prepare_read_bio(io->sbi->sb, fio->type, io->bio); |
6a8f8ca5 | 242 | else |
554b5125 | 243 | trace_f2fs_prepare_write_bio(io->sbi->sb, fio->type, io->bio); |
04d328de | 244 | |
4fc29c1a | 245 | __submit_bio(io->sbi, io->bio, fio->type); |
93dfe2ac JK |
246 | io->bio = NULL; |
247 | } | |
248 | ||
942fd319 JK |
249 | static bool __has_merged_page(struct f2fs_bio_info *io, |
250 | struct inode *inode, nid_t ino, pgoff_t idx) | |
0fd785eb | 251 | { |
0fd785eb CY |
252 | struct bio_vec *bvec; |
253 | struct page *target; | |
254 | int i; | |
255 | ||
0c3a5797 | 256 | if (!io->bio) |
0fd785eb | 257 | return false; |
0c3a5797 | 258 | |
942fd319 | 259 | if (!inode && !ino) |
0c3a5797 | 260 | return true; |
0fd785eb CY |
261 | |
262 | bio_for_each_segment_all(bvec, io->bio, i) { | |
263 | ||
0b81d077 | 264 | if (bvec->bv_page->mapping) |
0fd785eb | 265 | target = bvec->bv_page; |
0b81d077 JK |
266 | else |
267 | target = fscrypt_control_page(bvec->bv_page); | |
0fd785eb | 268 | |
942fd319 JK |
269 | if (idx != target->index) |
270 | continue; | |
271 | ||
0c3a5797 CY |
272 | if (inode && inode == target->mapping->host) |
273 | return true; | |
0c3a5797 | 274 | if (ino && ino == ino_of_node(target)) |
0fd785eb | 275 | return true; |
0fd785eb CY |
276 | } |
277 | ||
0fd785eb CY |
278 | return false; |
279 | } | |
280 | ||
0c3a5797 | 281 | static bool has_merged_page(struct f2fs_sb_info *sbi, struct inode *inode, |
942fd319 | 282 | nid_t ino, pgoff_t idx, enum page_type type) |
0c3a5797 CY |
283 | { |
284 | enum page_type btype = PAGE_TYPE_OF_BIO(type); | |
285 | struct f2fs_bio_info *io = &sbi->write_io[btype]; | |
286 | bool ret; | |
287 | ||
288 | down_read(&io->io_rwsem); | |
942fd319 | 289 | ret = __has_merged_page(io, inode, ino, idx); |
0c3a5797 CY |
290 | up_read(&io->io_rwsem); |
291 | return ret; | |
292 | } | |
293 | ||
294 | static void __f2fs_submit_merged_bio(struct f2fs_sb_info *sbi, | |
942fd319 JK |
295 | struct inode *inode, nid_t ino, pgoff_t idx, |
296 | enum page_type type, int rw) | |
93dfe2ac JK |
297 | { |
298 | enum page_type btype = PAGE_TYPE_OF_BIO(type); | |
299 | struct f2fs_bio_info *io; | |
300 | ||
301 | io = is_read_io(rw) ? &sbi->read_io : &sbi->write_io[btype]; | |
302 | ||
df0f8dc0 | 303 | down_write(&io->io_rwsem); |
458e6197 | 304 | |
942fd319 | 305 | if (!__has_merged_page(io, inode, ino, idx)) |
0c3a5797 CY |
306 | goto out; |
307 | ||
458e6197 JK |
308 | /* change META to META_FLUSH in the checkpoint procedure */ |
309 | if (type >= META_FLUSH) { | |
310 | io->fio.type = META_FLUSH; | |
04d328de | 311 | io->fio.op = REQ_OP_WRITE; |
7f54f51f | 312 | io->fio.op_flags = REQ_META | REQ_PRIO; |
70fd7614 | 313 | if (!test_opt(sbi, NOBARRIER)) |
7f54f51f | 314 | io->fio.op_flags |= REQ_PREFLUSH | REQ_FUA; |
458e6197 JK |
315 | } |
316 | __submit_merged_bio(io); | |
0c3a5797 | 317 | out: |
df0f8dc0 | 318 | up_write(&io->io_rwsem); |
93dfe2ac JK |
319 | } |
320 | ||
0c3a5797 CY |
321 | void f2fs_submit_merged_bio(struct f2fs_sb_info *sbi, enum page_type type, |
322 | int rw) | |
323 | { | |
942fd319 | 324 | __f2fs_submit_merged_bio(sbi, NULL, 0, 0, type, rw); |
0c3a5797 CY |
325 | } |
326 | ||
327 | void f2fs_submit_merged_bio_cond(struct f2fs_sb_info *sbi, | |
942fd319 JK |
328 | struct inode *inode, nid_t ino, pgoff_t idx, |
329 | enum page_type type, int rw) | |
0c3a5797 | 330 | { |
942fd319 JK |
331 | if (has_merged_page(sbi, inode, ino, idx, type)) |
332 | __f2fs_submit_merged_bio(sbi, inode, ino, idx, type, rw); | |
0c3a5797 CY |
333 | } |
334 | ||
406657dd CY |
335 | void f2fs_flush_merged_bios(struct f2fs_sb_info *sbi) |
336 | { | |
337 | f2fs_submit_merged_bio(sbi, DATA, WRITE); | |
338 | f2fs_submit_merged_bio(sbi, NODE, WRITE); | |
339 | f2fs_submit_merged_bio(sbi, META, WRITE); | |
340 | } | |
341 | ||
93dfe2ac JK |
342 | /* |
343 | * Fill the locked page with data located in the block address. | |
344 | * Return unlocked page. | |
345 | */ | |
05ca3632 | 346 | int f2fs_submit_page_bio(struct f2fs_io_info *fio) |
93dfe2ac | 347 | { |
93dfe2ac | 348 | struct bio *bio; |
0b81d077 JK |
349 | struct page *page = fio->encrypted_page ? |
350 | fio->encrypted_page : fio->page; | |
93dfe2ac | 351 | |
2ace38e0 | 352 | trace_f2fs_submit_page_bio(page, fio); |
05ca3632 | 353 | f2fs_trace_ios(fio, 0); |
93dfe2ac JK |
354 | |
355 | /* Allocate a new bio */ | |
04d328de | 356 | bio = __bio_alloc(fio->sbi, fio->new_blkaddr, 1, is_read_io(fio->op)); |
93dfe2ac | 357 | |
09cbfeaf | 358 | if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) { |
93dfe2ac | 359 | bio_put(bio); |
93dfe2ac JK |
360 | return -EFAULT; |
361 | } | |
04d328de | 362 | bio_set_op_attrs(bio, fio->op, fio->op_flags); |
93dfe2ac | 363 | |
4fc29c1a | 364 | __submit_bio(fio->sbi, bio, fio->type); |
93dfe2ac JK |
365 | return 0; |
366 | } | |
367 | ||
0a595eba | 368 | int f2fs_submit_page_mbio(struct f2fs_io_info *fio) |
93dfe2ac | 369 | { |
05ca3632 | 370 | struct f2fs_sb_info *sbi = fio->sbi; |
458e6197 | 371 | enum page_type btype = PAGE_TYPE_OF_BIO(fio->type); |
93dfe2ac | 372 | struct f2fs_bio_info *io; |
04d328de | 373 | bool is_read = is_read_io(fio->op); |
4375a336 | 374 | struct page *bio_page; |
0a595eba | 375 | int err = 0; |
93dfe2ac | 376 | |
940a6d34 | 377 | io = is_read ? &sbi->read_io : &sbi->write_io[btype]; |
93dfe2ac | 378 | |
7a9d7548 CY |
379 | if (fio->old_blkaddr != NEW_ADDR) |
380 | verify_block_addr(sbi, fio->old_blkaddr); | |
381 | verify_block_addr(sbi, fio->new_blkaddr); | |
93dfe2ac | 382 | |
36951b38 CY |
383 | bio_page = fio->encrypted_page ? fio->encrypted_page : fio->page; |
384 | ||
d68f735b JK |
385 | /* set submitted = 1 as a return value */ |
386 | fio->submitted = 1; | |
387 | ||
36951b38 CY |
388 | if (!is_read) |
389 | inc_page_count(sbi, WB_DATA_TYPE(bio_page)); | |
390 | ||
df0f8dc0 | 391 | down_write(&io->io_rwsem); |
93dfe2ac | 392 | |
7a9d7548 | 393 | if (io->bio && (io->last_block_in_bio != fio->new_blkaddr - 1 || |
3c62be17 JK |
394 | (io->fio.op != fio->op || io->fio.op_flags != fio->op_flags) || |
395 | !__same_bdev(sbi, fio->new_blkaddr, io->bio))) | |
458e6197 | 396 | __submit_merged_bio(io); |
93dfe2ac JK |
397 | alloc_new: |
398 | if (io->bio == NULL) { | |
0a595eba JK |
399 | if ((fio->type == DATA || fio->type == NODE) && |
400 | fio->new_blkaddr & F2FS_IO_SIZE_MASK(sbi)) { | |
401 | err = -EAGAIN; | |
540faedb CY |
402 | if (!is_read) |
403 | dec_page_count(sbi, WB_DATA_TYPE(bio_page)); | |
0a595eba JK |
404 | goto out_fail; |
405 | } | |
7a9d7548 | 406 | io->bio = __bio_alloc(sbi, fio->new_blkaddr, |
664ba972 | 407 | BIO_MAX_PAGES, is_read); |
458e6197 | 408 | io->fio = *fio; |
93dfe2ac JK |
409 | } |
410 | ||
09cbfeaf KS |
411 | if (bio_add_page(io->bio, bio_page, PAGE_SIZE, 0) < |
412 | PAGE_SIZE) { | |
458e6197 | 413 | __submit_merged_bio(io); |
93dfe2ac JK |
414 | goto alloc_new; |
415 | } | |
416 | ||
7a9d7548 | 417 | io->last_block_in_bio = fio->new_blkaddr; |
05ca3632 | 418 | f2fs_trace_ios(fio, 0); |
0a595eba | 419 | out_fail: |
df0f8dc0 | 420 | up_write(&io->io_rwsem); |
05ca3632 | 421 | trace_f2fs_submit_page_mbio(fio->page, fio); |
0a595eba | 422 | return err; |
93dfe2ac JK |
423 | } |
424 | ||
46008c6d CY |
425 | static void __set_data_blkaddr(struct dnode_of_data *dn) |
426 | { | |
427 | struct f2fs_node *rn = F2FS_NODE(dn->node_page); | |
428 | __le32 *addr_array; | |
429 | ||
430 | /* Get physical address of data block */ | |
431 | addr_array = blkaddr_in_node(rn); | |
432 | addr_array[dn->ofs_in_node] = cpu_to_le32(dn->data_blkaddr); | |
433 | } | |
434 | ||
0a8165d7 | 435 | /* |
eb47b800 JK |
436 | * Lock ordering for the change of data block address: |
437 | * ->data_page | |
438 | * ->node_page | |
439 | * update block addresses in the node page | |
440 | */ | |
216a620a | 441 | void set_data_blkaddr(struct dnode_of_data *dn) |
eb47b800 | 442 | { |
46008c6d CY |
443 | f2fs_wait_on_page_writeback(dn->node_page, NODE, true); |
444 | __set_data_blkaddr(dn); | |
445 | if (set_page_dirty(dn->node_page)) | |
12719ae1 | 446 | dn->node_changed = true; |
eb47b800 JK |
447 | } |
448 | ||
f28b3434 CY |
449 | void f2fs_update_data_blkaddr(struct dnode_of_data *dn, block_t blkaddr) |
450 | { | |
451 | dn->data_blkaddr = blkaddr; | |
452 | set_data_blkaddr(dn); | |
453 | f2fs_update_extent_cache(dn); | |
454 | } | |
455 | ||
46008c6d CY |
456 | /* dn->ofs_in_node will be returned with up-to-date last block pointer */ |
457 | int reserve_new_blocks(struct dnode_of_data *dn, blkcnt_t count) | |
eb47b800 | 458 | { |
4081363f | 459 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
eb47b800 | 460 | |
46008c6d CY |
461 | if (!count) |
462 | return 0; | |
463 | ||
91942321 | 464 | if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC))) |
eb47b800 | 465 | return -EPERM; |
46008c6d | 466 | if (unlikely(!inc_valid_block_count(sbi, dn->inode, &count))) |
eb47b800 JK |
467 | return -ENOSPC; |
468 | ||
46008c6d CY |
469 | trace_f2fs_reserve_new_blocks(dn->inode, dn->nid, |
470 | dn->ofs_in_node, count); | |
471 | ||
472 | f2fs_wait_on_page_writeback(dn->node_page, NODE, true); | |
473 | ||
474 | for (; count > 0; dn->ofs_in_node++) { | |
475 | block_t blkaddr = | |
476 | datablock_addr(dn->node_page, dn->ofs_in_node); | |
477 | if (blkaddr == NULL_ADDR) { | |
478 | dn->data_blkaddr = NEW_ADDR; | |
479 | __set_data_blkaddr(dn); | |
480 | count--; | |
481 | } | |
482 | } | |
483 | ||
484 | if (set_page_dirty(dn->node_page)) | |
485 | dn->node_changed = true; | |
eb47b800 JK |
486 | return 0; |
487 | } | |
488 | ||
46008c6d CY |
489 | /* Should keep dn->ofs_in_node unchanged */ |
490 | int reserve_new_block(struct dnode_of_data *dn) | |
491 | { | |
492 | unsigned int ofs_in_node = dn->ofs_in_node; | |
493 | int ret; | |
494 | ||
495 | ret = reserve_new_blocks(dn, 1); | |
496 | dn->ofs_in_node = ofs_in_node; | |
497 | return ret; | |
498 | } | |
499 | ||
b600965c HL |
500 | int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index) |
501 | { | |
502 | bool need_put = dn->inode_page ? false : true; | |
503 | int err; | |
504 | ||
505 | err = get_dnode_of_data(dn, index, ALLOC_NODE); | |
506 | if (err) | |
507 | return err; | |
a8865372 | 508 | |
b600965c HL |
509 | if (dn->data_blkaddr == NULL_ADDR) |
510 | err = reserve_new_block(dn); | |
a8865372 | 511 | if (err || need_put) |
b600965c HL |
512 | f2fs_put_dnode(dn); |
513 | return err; | |
514 | } | |
515 | ||
759af1c9 | 516 | int f2fs_get_block(struct dnode_of_data *dn, pgoff_t index) |
eb47b800 | 517 | { |
e15882b6 | 518 | struct extent_info ei = {0,0,0}; |
759af1c9 | 519 | struct inode *inode = dn->inode; |
028a41e8 | 520 | |
759af1c9 FL |
521 | if (f2fs_lookup_extent_cache(inode, index, &ei)) { |
522 | dn->data_blkaddr = ei.blk + index - ei.fofs; | |
523 | return 0; | |
429511cd | 524 | } |
028a41e8 | 525 | |
759af1c9 | 526 | return f2fs_reserve_block(dn, index); |
eb47b800 JK |
527 | } |
528 | ||
a56c7c6f | 529 | struct page *get_read_data_page(struct inode *inode, pgoff_t index, |
04d328de | 530 | int op_flags, bool for_write) |
eb47b800 | 531 | { |
eb47b800 JK |
532 | struct address_space *mapping = inode->i_mapping; |
533 | struct dnode_of_data dn; | |
534 | struct page *page; | |
e15882b6 | 535 | struct extent_info ei = {0,0,0}; |
eb47b800 | 536 | int err; |
cf04e8eb | 537 | struct f2fs_io_info fio = { |
05ca3632 | 538 | .sbi = F2FS_I_SB(inode), |
cf04e8eb | 539 | .type = DATA, |
04d328de MC |
540 | .op = REQ_OP_READ, |
541 | .op_flags = op_flags, | |
4375a336 | 542 | .encrypted_page = NULL, |
cf04e8eb | 543 | }; |
eb47b800 | 544 | |
4375a336 JK |
545 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) |
546 | return read_mapping_page(mapping, index, NULL); | |
547 | ||
a56c7c6f | 548 | page = f2fs_grab_cache_page(mapping, index, for_write); |
650495de JK |
549 | if (!page) |
550 | return ERR_PTR(-ENOMEM); | |
551 | ||
cb3bc9ee CY |
552 | if (f2fs_lookup_extent_cache(inode, index, &ei)) { |
553 | dn.data_blkaddr = ei.blk + index - ei.fofs; | |
554 | goto got_it; | |
555 | } | |
556 | ||
eb47b800 | 557 | set_new_dnode(&dn, inode, NULL, NULL, 0); |
266e97a8 | 558 | err = get_dnode_of_data(&dn, index, LOOKUP_NODE); |
86531d6b JK |
559 | if (err) |
560 | goto put_err; | |
eb47b800 JK |
561 | f2fs_put_dnode(&dn); |
562 | ||
6bacf52f | 563 | if (unlikely(dn.data_blkaddr == NULL_ADDR)) { |
86531d6b JK |
564 | err = -ENOENT; |
565 | goto put_err; | |
650495de | 566 | } |
cb3bc9ee | 567 | got_it: |
43f3eae1 JK |
568 | if (PageUptodate(page)) { |
569 | unlock_page(page); | |
eb47b800 | 570 | return page; |
43f3eae1 | 571 | } |
eb47b800 | 572 | |
d59ff4df JK |
573 | /* |
574 | * A new dentry page is allocated but not able to be written, since its | |
575 | * new inode page couldn't be allocated due to -ENOSPC. | |
576 | * In such the case, its blkaddr can be remained as NEW_ADDR. | |
577 | * see, f2fs_add_link -> get_new_data_page -> init_inode_metadata. | |
578 | */ | |
579 | if (dn.data_blkaddr == NEW_ADDR) { | |
09cbfeaf | 580 | zero_user_segment(page, 0, PAGE_SIZE); |
237c0790 JK |
581 | if (!PageUptodate(page)) |
582 | SetPageUptodate(page); | |
43f3eae1 | 583 | unlock_page(page); |
d59ff4df JK |
584 | return page; |
585 | } | |
eb47b800 | 586 | |
7a9d7548 | 587 | fio.new_blkaddr = fio.old_blkaddr = dn.data_blkaddr; |
05ca3632 JK |
588 | fio.page = page; |
589 | err = f2fs_submit_page_bio(&fio); | |
393ff91f | 590 | if (err) |
86531d6b | 591 | goto put_err; |
43f3eae1 | 592 | return page; |
86531d6b JK |
593 | |
594 | put_err: | |
595 | f2fs_put_page(page, 1); | |
596 | return ERR_PTR(err); | |
43f3eae1 JK |
597 | } |
598 | ||
599 | struct page *find_data_page(struct inode *inode, pgoff_t index) | |
600 | { | |
601 | struct address_space *mapping = inode->i_mapping; | |
602 | struct page *page; | |
603 | ||
604 | page = find_get_page(mapping, index); | |
605 | if (page && PageUptodate(page)) | |
606 | return page; | |
607 | f2fs_put_page(page, 0); | |
608 | ||
70fd7614 | 609 | page = get_read_data_page(inode, index, 0, false); |
43f3eae1 JK |
610 | if (IS_ERR(page)) |
611 | return page; | |
612 | ||
613 | if (PageUptodate(page)) | |
614 | return page; | |
615 | ||
616 | wait_on_page_locked(page); | |
617 | if (unlikely(!PageUptodate(page))) { | |
618 | f2fs_put_page(page, 0); | |
619 | return ERR_PTR(-EIO); | |
620 | } | |
621 | return page; | |
622 | } | |
623 | ||
624 | /* | |
625 | * If it tries to access a hole, return an error. | |
626 | * Because, the callers, functions in dir.c and GC, should be able to know | |
627 | * whether this page exists or not. | |
628 | */ | |
a56c7c6f JK |
629 | struct page *get_lock_data_page(struct inode *inode, pgoff_t index, |
630 | bool for_write) | |
43f3eae1 JK |
631 | { |
632 | struct address_space *mapping = inode->i_mapping; | |
633 | struct page *page; | |
634 | repeat: | |
70fd7614 | 635 | page = get_read_data_page(inode, index, 0, for_write); |
43f3eae1 JK |
636 | if (IS_ERR(page)) |
637 | return page; | |
393ff91f | 638 | |
43f3eae1 | 639 | /* wait for read completion */ |
393ff91f | 640 | lock_page(page); |
6bacf52f | 641 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
642 | f2fs_put_page(page, 1); |
643 | goto repeat; | |
eb47b800 | 644 | } |
1563ac75 CY |
645 | if (unlikely(!PageUptodate(page))) { |
646 | f2fs_put_page(page, 1); | |
647 | return ERR_PTR(-EIO); | |
648 | } | |
eb47b800 JK |
649 | return page; |
650 | } | |
651 | ||
0a8165d7 | 652 | /* |
eb47b800 JK |
653 | * Caller ensures that this data page is never allocated. |
654 | * A new zero-filled data page is allocated in the page cache. | |
39936837 | 655 | * |
4f4124d0 CY |
656 | * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and |
657 | * f2fs_unlock_op(). | |
470f00e9 CY |
658 | * Note that, ipage is set only by make_empty_dir, and if any error occur, |
659 | * ipage should be released by this function. | |
eb47b800 | 660 | */ |
64aa7ed9 | 661 | struct page *get_new_data_page(struct inode *inode, |
a8865372 | 662 | struct page *ipage, pgoff_t index, bool new_i_size) |
eb47b800 | 663 | { |
eb47b800 JK |
664 | struct address_space *mapping = inode->i_mapping; |
665 | struct page *page; | |
666 | struct dnode_of_data dn; | |
667 | int err; | |
7612118a | 668 | |
a56c7c6f | 669 | page = f2fs_grab_cache_page(mapping, index, true); |
470f00e9 CY |
670 | if (!page) { |
671 | /* | |
672 | * before exiting, we should make sure ipage will be released | |
673 | * if any error occur. | |
674 | */ | |
675 | f2fs_put_page(ipage, 1); | |
01f28610 | 676 | return ERR_PTR(-ENOMEM); |
470f00e9 | 677 | } |
eb47b800 | 678 | |
a8865372 | 679 | set_new_dnode(&dn, inode, ipage, NULL, 0); |
b600965c | 680 | err = f2fs_reserve_block(&dn, index); |
01f28610 JK |
681 | if (err) { |
682 | f2fs_put_page(page, 1); | |
eb47b800 | 683 | return ERR_PTR(err); |
a8865372 | 684 | } |
01f28610 JK |
685 | if (!ipage) |
686 | f2fs_put_dnode(&dn); | |
eb47b800 JK |
687 | |
688 | if (PageUptodate(page)) | |
01f28610 | 689 | goto got_it; |
eb47b800 JK |
690 | |
691 | if (dn.data_blkaddr == NEW_ADDR) { | |
09cbfeaf | 692 | zero_user_segment(page, 0, PAGE_SIZE); |
237c0790 JK |
693 | if (!PageUptodate(page)) |
694 | SetPageUptodate(page); | |
eb47b800 | 695 | } else { |
4375a336 | 696 | f2fs_put_page(page, 1); |
a8865372 | 697 | |
7612118a JK |
698 | /* if ipage exists, blkaddr should be NEW_ADDR */ |
699 | f2fs_bug_on(F2FS_I_SB(inode), ipage); | |
700 | page = get_lock_data_page(inode, index, true); | |
4375a336 | 701 | if (IS_ERR(page)) |
7612118a | 702 | return page; |
eb47b800 | 703 | } |
01f28610 | 704 | got_it: |
9edcdabf | 705 | if (new_i_size && i_size_read(inode) < |
ee6d182f | 706 | ((loff_t)(index + 1) << PAGE_SHIFT)) |
fc9581c8 | 707 | f2fs_i_size_write(inode, ((loff_t)(index + 1) << PAGE_SHIFT)); |
eb47b800 JK |
708 | return page; |
709 | } | |
710 | ||
bfad7c2d JK |
711 | static int __allocate_data_block(struct dnode_of_data *dn) |
712 | { | |
4081363f | 713 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
bfad7c2d | 714 | struct f2fs_summary sum; |
bfad7c2d | 715 | struct node_info ni; |
976e4c50 | 716 | pgoff_t fofs; |
46008c6d | 717 | blkcnt_t count = 1; |
bfad7c2d | 718 | |
91942321 | 719 | if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC))) |
bfad7c2d | 720 | return -EPERM; |
df6136ef CY |
721 | |
722 | dn->data_blkaddr = datablock_addr(dn->node_page, dn->ofs_in_node); | |
723 | if (dn->data_blkaddr == NEW_ADDR) | |
724 | goto alloc; | |
725 | ||
46008c6d | 726 | if (unlikely(!inc_valid_block_count(sbi, dn->inode, &count))) |
bfad7c2d JK |
727 | return -ENOSPC; |
728 | ||
df6136ef | 729 | alloc: |
bfad7c2d JK |
730 | get_node_info(sbi, dn->nid, &ni); |
731 | set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version); | |
732 | ||
df6136ef | 733 | allocate_data_block(sbi, NULL, dn->data_blkaddr, &dn->data_blkaddr, |
6ae1be13 | 734 | &sum, CURSEG_WARM_DATA); |
216a620a | 735 | set_data_blkaddr(dn); |
bfad7c2d | 736 | |
976e4c50 | 737 | /* update i_size */ |
81ca7350 | 738 | fofs = start_bidx_of_node(ofs_of_node(dn->node_page), dn->inode) + |
976e4c50 | 739 | dn->ofs_in_node; |
09cbfeaf | 740 | if (i_size_read(dn->inode) < ((loff_t)(fofs + 1) << PAGE_SHIFT)) |
fc9581c8 | 741 | f2fs_i_size_write(dn->inode, |
09cbfeaf | 742 | ((loff_t)(fofs + 1) << PAGE_SHIFT)); |
bfad7c2d JK |
743 | return 0; |
744 | } | |
745 | ||
c040ff9d JK |
746 | static inline bool __force_buffered_io(struct inode *inode, int rw) |
747 | { | |
748 | return ((f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) || | |
749 | (rw == WRITE && test_opt(F2FS_I_SB(inode), LFS)) || | |
750 | F2FS_I_SB(inode)->s_ndevs); | |
751 | } | |
752 | ||
a7de6086 | 753 | int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *from) |
59b802e5 | 754 | { |
b439b103 | 755 | struct inode *inode = file_inode(iocb->ki_filp); |
5b8db7fa | 756 | struct f2fs_map_blocks map; |
a7de6086 | 757 | int err = 0; |
59b802e5 | 758 | |
dc91de78 JK |
759 | if (is_inode_flag_set(inode, FI_NO_PREALLOC)) |
760 | return 0; | |
761 | ||
0080c507 | 762 | map.m_lblk = F2FS_BLK_ALIGN(iocb->ki_pos); |
dfd02e4d CY |
763 | map.m_len = F2FS_BYTES_TO_BLK(iocb->ki_pos + iov_iter_count(from)); |
764 | if (map.m_len > map.m_lblk) | |
765 | map.m_len -= map.m_lblk; | |
766 | else | |
767 | map.m_len = 0; | |
768 | ||
da85985c | 769 | map.m_next_pgofs = NULL; |
2a340760 | 770 | |
24b84912 | 771 | if (iocb->ki_flags & IOCB_DIRECT) { |
a7de6086 JK |
772 | err = f2fs_convert_inline_inode(inode); |
773 | if (err) | |
774 | return err; | |
c040ff9d JK |
775 | return f2fs_map_blocks(inode, &map, 1, |
776 | __force_buffered_io(inode, WRITE) ? | |
777 | F2FS_GET_BLOCK_PRE_AIO : | |
778 | F2FS_GET_BLOCK_PRE_DIO); | |
24b84912 JK |
779 | } |
780 | if (iocb->ki_pos + iov_iter_count(from) > MAX_INLINE_DATA) { | |
a7de6086 JK |
781 | err = f2fs_convert_inline_inode(inode); |
782 | if (err) | |
783 | return err; | |
b439b103 | 784 | } |
24b84912 JK |
785 | if (!f2fs_has_inline_data(inode)) |
786 | return f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO); | |
a7de6086 | 787 | return err; |
59b802e5 JK |
788 | } |
789 | ||
59c9081b YH |
790 | static inline void __do_map_lock(struct f2fs_sb_info *sbi, int flag, bool lock) |
791 | { | |
792 | if (flag == F2FS_GET_BLOCK_PRE_AIO) { | |
793 | if (lock) | |
794 | down_read(&sbi->node_change); | |
795 | else | |
796 | up_read(&sbi->node_change); | |
797 | } else { | |
798 | if (lock) | |
799 | f2fs_lock_op(sbi); | |
800 | else | |
801 | f2fs_unlock_op(sbi); | |
802 | } | |
803 | } | |
804 | ||
0a8165d7 | 805 | /* |
003a3e1d JK |
806 | * f2fs_map_blocks() now supported readahead/bmap/rw direct_IO with |
807 | * f2fs_map_blocks structure. | |
4f4124d0 CY |
808 | * If original data blocks are allocated, then give them to blockdev. |
809 | * Otherwise, | |
810 | * a. preallocate requested block addresses | |
811 | * b. do not use extent cache for better performance | |
812 | * c. give the block addresses to blockdev | |
eb47b800 | 813 | */ |
d323d005 | 814 | int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map, |
e2b4e2bc | 815 | int create, int flag) |
eb47b800 | 816 | { |
003a3e1d | 817 | unsigned int maxblocks = map->m_len; |
eb47b800 | 818 | struct dnode_of_data dn; |
f9811703 | 819 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
ac6f1999 | 820 | int mode = create ? ALLOC_NODE : LOOKUP_NODE; |
46008c6d | 821 | pgoff_t pgofs, end_offset, end; |
bfad7c2d | 822 | int err = 0, ofs = 1; |
46008c6d CY |
823 | unsigned int ofs_in_node, last_ofs_in_node; |
824 | blkcnt_t prealloc; | |
e15882b6 | 825 | struct extent_info ei = {0,0,0}; |
7df3a431 | 826 | block_t blkaddr; |
eb47b800 | 827 | |
dfd02e4d CY |
828 | if (!maxblocks) |
829 | return 0; | |
830 | ||
003a3e1d JK |
831 | map->m_len = 0; |
832 | map->m_flags = 0; | |
833 | ||
834 | /* it only supports block size == page size */ | |
835 | pgofs = (pgoff_t)map->m_lblk; | |
46008c6d | 836 | end = pgofs + maxblocks; |
eb47b800 | 837 | |
24b84912 | 838 | if (!create && f2fs_lookup_extent_cache(inode, pgofs, &ei)) { |
003a3e1d JK |
839 | map->m_pblk = ei.blk + pgofs - ei.fofs; |
840 | map->m_len = min((pgoff_t)maxblocks, ei.fofs + ei.len - pgofs); | |
841 | map->m_flags = F2FS_MAP_MAPPED; | |
bfad7c2d | 842 | goto out; |
a2e7d1bf | 843 | } |
bfad7c2d | 844 | |
4fe71e88 | 845 | next_dnode: |
59b802e5 | 846 | if (create) |
59c9081b | 847 | __do_map_lock(sbi, flag, true); |
eb47b800 JK |
848 | |
849 | /* When reading holes, we need its node page */ | |
850 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
bfad7c2d | 851 | err = get_dnode_of_data(&dn, pgofs, mode); |
1ec79083 | 852 | if (err) { |
43473f96 CY |
853 | if (flag == F2FS_GET_BLOCK_BMAP) |
854 | map->m_pblk = 0; | |
da85985c | 855 | if (err == -ENOENT) { |
bfad7c2d | 856 | err = 0; |
da85985c CY |
857 | if (map->m_next_pgofs) |
858 | *map->m_next_pgofs = | |
859 | get_next_page_offset(&dn, pgofs); | |
860 | } | |
bfad7c2d | 861 | goto unlock_out; |
848753aa | 862 | } |
973163fc | 863 | |
46008c6d | 864 | prealloc = 0; |
230436b3 | 865 | last_ofs_in_node = ofs_in_node = dn.ofs_in_node; |
81ca7350 | 866 | end_offset = ADDRS_PER_PAGE(dn.node_page, inode); |
4fe71e88 CY |
867 | |
868 | next_block: | |
869 | blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node); | |
870 | ||
871 | if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR) { | |
973163fc | 872 | if (create) { |
f9811703 CY |
873 | if (unlikely(f2fs_cp_error(sbi))) { |
874 | err = -EIO; | |
4fe71e88 | 875 | goto sync_out; |
f9811703 | 876 | } |
24b84912 | 877 | if (flag == F2FS_GET_BLOCK_PRE_AIO) { |
46008c6d CY |
878 | if (blkaddr == NULL_ADDR) { |
879 | prealloc++; | |
880 | last_ofs_in_node = dn.ofs_in_node; | |
881 | } | |
24b84912 JK |
882 | } else { |
883 | err = __allocate_data_block(&dn); | |
6f2d8ed6 | 884 | if (!err) |
91942321 | 885 | set_inode_flag(inode, FI_APPEND_WRITE); |
24b84912 | 886 | } |
973163fc | 887 | if (err) |
4fe71e88 | 888 | goto sync_out; |
3f2be043 | 889 | map->m_flags |= F2FS_MAP_NEW; |
4fe71e88 | 890 | blkaddr = dn.data_blkaddr; |
973163fc | 891 | } else { |
43473f96 CY |
892 | if (flag == F2FS_GET_BLOCK_BMAP) { |
893 | map->m_pblk = 0; | |
894 | goto sync_out; | |
895 | } | |
da85985c CY |
896 | if (flag == F2FS_GET_BLOCK_FIEMAP && |
897 | blkaddr == NULL_ADDR) { | |
898 | if (map->m_next_pgofs) | |
899 | *map->m_next_pgofs = pgofs + 1; | |
900 | } | |
973163fc | 901 | if (flag != F2FS_GET_BLOCK_FIEMAP || |
43473f96 | 902 | blkaddr != NEW_ADDR) |
4fe71e88 | 903 | goto sync_out; |
e2b4e2bc | 904 | } |
e2b4e2bc | 905 | } |
eb47b800 | 906 | |
46008c6d CY |
907 | if (flag == F2FS_GET_BLOCK_PRE_AIO) |
908 | goto skip; | |
909 | ||
4fe71e88 CY |
910 | if (map->m_len == 0) { |
911 | /* preallocated unwritten block should be mapped for fiemap. */ | |
912 | if (blkaddr == NEW_ADDR) | |
913 | map->m_flags |= F2FS_MAP_UNWRITTEN; | |
914 | map->m_flags |= F2FS_MAP_MAPPED; | |
915 | ||
916 | map->m_pblk = blkaddr; | |
917 | map->m_len = 1; | |
918 | } else if ((map->m_pblk != NEW_ADDR && | |
919 | blkaddr == (map->m_pblk + ofs)) || | |
b439b103 | 920 | (map->m_pblk == NEW_ADDR && blkaddr == NEW_ADDR) || |
46008c6d | 921 | flag == F2FS_GET_BLOCK_PRE_DIO) { |
4fe71e88 CY |
922 | ofs++; |
923 | map->m_len++; | |
924 | } else { | |
925 | goto sync_out; | |
926 | } | |
bfad7c2d | 927 | |
46008c6d | 928 | skip: |
bfad7c2d JK |
929 | dn.ofs_in_node++; |
930 | pgofs++; | |
931 | ||
46008c6d CY |
932 | /* preallocate blocks in batch for one dnode page */ |
933 | if (flag == F2FS_GET_BLOCK_PRE_AIO && | |
934 | (pgofs == end || dn.ofs_in_node == end_offset)) { | |
7df3a431 | 935 | |
46008c6d CY |
936 | dn.ofs_in_node = ofs_in_node; |
937 | err = reserve_new_blocks(&dn, prealloc); | |
938 | if (err) | |
939 | goto sync_out; | |
bfad7c2d | 940 | |
46008c6d CY |
941 | map->m_len += dn.ofs_in_node - ofs_in_node; |
942 | if (prealloc && dn.ofs_in_node != last_ofs_in_node + 1) { | |
943 | err = -ENOSPC; | |
944 | goto sync_out; | |
3104af35 | 945 | } |
46008c6d CY |
946 | dn.ofs_in_node = end_offset; |
947 | } | |
948 | ||
949 | if (pgofs >= end) | |
950 | goto sync_out; | |
951 | else if (dn.ofs_in_node < end_offset) | |
952 | goto next_block; | |
953 | ||
46008c6d CY |
954 | f2fs_put_dnode(&dn); |
955 | ||
956 | if (create) { | |
59c9081b | 957 | __do_map_lock(sbi, flag, false); |
6f2d8ed6 | 958 | f2fs_balance_fs(sbi, dn.node_changed); |
eb47b800 | 959 | } |
46008c6d | 960 | goto next_dnode; |
7df3a431 | 961 | |
bfad7c2d | 962 | sync_out: |
eb47b800 | 963 | f2fs_put_dnode(&dn); |
bfad7c2d | 964 | unlock_out: |
2a340760 | 965 | if (create) { |
59c9081b | 966 | __do_map_lock(sbi, flag, false); |
6f2d8ed6 | 967 | f2fs_balance_fs(sbi, dn.node_changed); |
2a340760 | 968 | } |
bfad7c2d | 969 | out: |
003a3e1d | 970 | trace_f2fs_map_blocks(inode, map, err); |
bfad7c2d | 971 | return err; |
eb47b800 JK |
972 | } |
973 | ||
003a3e1d | 974 | static int __get_data_block(struct inode *inode, sector_t iblock, |
da85985c CY |
975 | struct buffer_head *bh, int create, int flag, |
976 | pgoff_t *next_pgofs) | |
003a3e1d JK |
977 | { |
978 | struct f2fs_map_blocks map; | |
a7de6086 | 979 | int err; |
003a3e1d JK |
980 | |
981 | map.m_lblk = iblock; | |
982 | map.m_len = bh->b_size >> inode->i_blkbits; | |
da85985c | 983 | map.m_next_pgofs = next_pgofs; |
003a3e1d | 984 | |
a7de6086 JK |
985 | err = f2fs_map_blocks(inode, &map, create, flag); |
986 | if (!err) { | |
003a3e1d JK |
987 | map_bh(bh, inode->i_sb, map.m_pblk); |
988 | bh->b_state = (bh->b_state & ~F2FS_MAP_FLAGS) | map.m_flags; | |
b86e3307 | 989 | bh->b_size = (u64)map.m_len << inode->i_blkbits; |
003a3e1d | 990 | } |
a7de6086 | 991 | return err; |
003a3e1d JK |
992 | } |
993 | ||
ccfb3000 | 994 | static int get_data_block(struct inode *inode, sector_t iblock, |
da85985c CY |
995 | struct buffer_head *bh_result, int create, int flag, |
996 | pgoff_t *next_pgofs) | |
e2b4e2bc | 997 | { |
da85985c CY |
998 | return __get_data_block(inode, iblock, bh_result, create, |
999 | flag, next_pgofs); | |
e2b4e2bc CY |
1000 | } |
1001 | ||
1002 | static int get_data_block_dio(struct inode *inode, sector_t iblock, | |
ccfb3000 JK |
1003 | struct buffer_head *bh_result, int create) |
1004 | { | |
e2b4e2bc | 1005 | return __get_data_block(inode, iblock, bh_result, create, |
da85985c | 1006 | F2FS_GET_BLOCK_DIO, NULL); |
ccfb3000 JK |
1007 | } |
1008 | ||
e2b4e2bc | 1009 | static int get_data_block_bmap(struct inode *inode, sector_t iblock, |
ccfb3000 JK |
1010 | struct buffer_head *bh_result, int create) |
1011 | { | |
179448bf | 1012 | /* Block number less than F2FS MAX BLOCKS */ |
e0afc4d6 | 1013 | if (unlikely(iblock >= F2FS_I_SB(inode)->max_file_blocks)) |
179448bf YH |
1014 | return -EFBIG; |
1015 | ||
e2b4e2bc | 1016 | return __get_data_block(inode, iblock, bh_result, create, |
da85985c | 1017 | F2FS_GET_BLOCK_BMAP, NULL); |
ccfb3000 JK |
1018 | } |
1019 | ||
7f63eb77 JK |
1020 | static inline sector_t logical_to_blk(struct inode *inode, loff_t offset) |
1021 | { | |
1022 | return (offset >> inode->i_blkbits); | |
1023 | } | |
1024 | ||
1025 | static inline loff_t blk_to_logical(struct inode *inode, sector_t blk) | |
1026 | { | |
1027 | return (blk << inode->i_blkbits); | |
1028 | } | |
1029 | ||
9ab70134 JK |
1030 | int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, |
1031 | u64 start, u64 len) | |
1032 | { | |
7f63eb77 JK |
1033 | struct buffer_head map_bh; |
1034 | sector_t start_blk, last_blk; | |
da85985c | 1035 | pgoff_t next_pgofs; |
7f63eb77 JK |
1036 | u64 logical = 0, phys = 0, size = 0; |
1037 | u32 flags = 0; | |
7f63eb77 JK |
1038 | int ret = 0; |
1039 | ||
1040 | ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC); | |
1041 | if (ret) | |
1042 | return ret; | |
1043 | ||
67f8cf3c JK |
1044 | if (f2fs_has_inline_data(inode)) { |
1045 | ret = f2fs_inline_data_fiemap(inode, fieinfo, start, len); | |
1046 | if (ret != -EAGAIN) | |
1047 | return ret; | |
1048 | } | |
1049 | ||
5955102c | 1050 | inode_lock(inode); |
de1475cc | 1051 | |
7f63eb77 JK |
1052 | if (logical_to_blk(inode, len) == 0) |
1053 | len = blk_to_logical(inode, 1); | |
1054 | ||
1055 | start_blk = logical_to_blk(inode, start); | |
1056 | last_blk = logical_to_blk(inode, start + len - 1); | |
9a950d52 | 1057 | |
7f63eb77 JK |
1058 | next: |
1059 | memset(&map_bh, 0, sizeof(struct buffer_head)); | |
1060 | map_bh.b_size = len; | |
1061 | ||
e2b4e2bc | 1062 | ret = get_data_block(inode, start_blk, &map_bh, 0, |
da85985c | 1063 | F2FS_GET_BLOCK_FIEMAP, &next_pgofs); |
7f63eb77 JK |
1064 | if (ret) |
1065 | goto out; | |
1066 | ||
1067 | /* HOLE */ | |
1068 | if (!buffer_mapped(&map_bh)) { | |
da85985c | 1069 | start_blk = next_pgofs; |
58736fa6 CY |
1070 | |
1071 | if (blk_to_logical(inode, start_blk) < blk_to_logical(inode, | |
1072 | F2FS_I_SB(inode)->max_file_blocks)) | |
9a950d52 | 1073 | goto prep_next; |
58736fa6 | 1074 | |
9a950d52 FL |
1075 | flags |= FIEMAP_EXTENT_LAST; |
1076 | } | |
7f63eb77 | 1077 | |
da5af127 CY |
1078 | if (size) { |
1079 | if (f2fs_encrypted_inode(inode)) | |
1080 | flags |= FIEMAP_EXTENT_DATA_ENCRYPTED; | |
1081 | ||
9a950d52 FL |
1082 | ret = fiemap_fill_next_extent(fieinfo, logical, |
1083 | phys, size, flags); | |
da5af127 | 1084 | } |
7f63eb77 | 1085 | |
9a950d52 FL |
1086 | if (start_blk > last_blk || ret) |
1087 | goto out; | |
7f63eb77 | 1088 | |
9a950d52 FL |
1089 | logical = blk_to_logical(inode, start_blk); |
1090 | phys = blk_to_logical(inode, map_bh.b_blocknr); | |
1091 | size = map_bh.b_size; | |
1092 | flags = 0; | |
1093 | if (buffer_unwritten(&map_bh)) | |
1094 | flags = FIEMAP_EXTENT_UNWRITTEN; | |
7f63eb77 | 1095 | |
9a950d52 | 1096 | start_blk += logical_to_blk(inode, size); |
7f63eb77 | 1097 | |
9a950d52 | 1098 | prep_next: |
7f63eb77 JK |
1099 | cond_resched(); |
1100 | if (fatal_signal_pending(current)) | |
1101 | ret = -EINTR; | |
1102 | else | |
1103 | goto next; | |
1104 | out: | |
1105 | if (ret == 1) | |
1106 | ret = 0; | |
1107 | ||
5955102c | 1108 | inode_unlock(inode); |
7f63eb77 | 1109 | return ret; |
9ab70134 JK |
1110 | } |
1111 | ||
6a7a3aed WY |
1112 | static struct bio *f2fs_grab_bio(struct inode *inode, block_t blkaddr, |
1113 | unsigned nr_pages) | |
78682f79 CY |
1114 | { |
1115 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); | |
1116 | struct fscrypt_ctx *ctx = NULL; | |
78682f79 CY |
1117 | struct bio *bio; |
1118 | ||
1119 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) { | |
1120 | ctx = fscrypt_get_ctx(inode, GFP_NOFS); | |
1121 | if (IS_ERR(ctx)) | |
1122 | return ERR_CAST(ctx); | |
1123 | ||
1124 | /* wait the page to be moved by cleaning */ | |
1125 | f2fs_wait_on_encrypted_page_writeback(sbi, blkaddr); | |
1126 | } | |
1127 | ||
1128 | bio = bio_alloc(GFP_KERNEL, min_t(int, nr_pages, BIO_MAX_PAGES)); | |
1129 | if (!bio) { | |
1130 | if (ctx) | |
1131 | fscrypt_release_ctx(ctx); | |
1132 | return ERR_PTR(-ENOMEM); | |
1133 | } | |
3c62be17 | 1134 | f2fs_target_device(sbi, blkaddr, bio); |
78682f79 CY |
1135 | bio->bi_end_io = f2fs_read_end_io; |
1136 | bio->bi_private = ctx; | |
1137 | ||
1138 | return bio; | |
1139 | } | |
1140 | ||
f1e88660 JK |
1141 | /* |
1142 | * This function was originally taken from fs/mpage.c, and customized for f2fs. | |
1143 | * Major change was from block_size == page_size in f2fs by default. | |
1144 | */ | |
1145 | static int f2fs_mpage_readpages(struct address_space *mapping, | |
1146 | struct list_head *pages, struct page *page, | |
1147 | unsigned nr_pages) | |
1148 | { | |
1149 | struct bio *bio = NULL; | |
1150 | unsigned page_idx; | |
1151 | sector_t last_block_in_bio = 0; | |
1152 | struct inode *inode = mapping->host; | |
1153 | const unsigned blkbits = inode->i_blkbits; | |
1154 | const unsigned blocksize = 1 << blkbits; | |
1155 | sector_t block_in_file; | |
1156 | sector_t last_block; | |
1157 | sector_t last_block_in_file; | |
1158 | sector_t block_nr; | |
f1e88660 JK |
1159 | struct f2fs_map_blocks map; |
1160 | ||
1161 | map.m_pblk = 0; | |
1162 | map.m_lblk = 0; | |
1163 | map.m_len = 0; | |
1164 | map.m_flags = 0; | |
da85985c | 1165 | map.m_next_pgofs = NULL; |
f1e88660 JK |
1166 | |
1167 | for (page_idx = 0; nr_pages; page_idx++, nr_pages--) { | |
1168 | ||
f1e88660 | 1169 | if (pages) { |
939afa94 | 1170 | page = list_last_entry(pages, struct page, lru); |
a83d50bc KM |
1171 | |
1172 | prefetchw(&page->flags); | |
f1e88660 JK |
1173 | list_del(&page->lru); |
1174 | if (add_to_page_cache_lru(page, mapping, | |
8a5c743e MH |
1175 | page->index, |
1176 | readahead_gfp_mask(mapping))) | |
f1e88660 JK |
1177 | goto next_page; |
1178 | } | |
1179 | ||
1180 | block_in_file = (sector_t)page->index; | |
1181 | last_block = block_in_file + nr_pages; | |
1182 | last_block_in_file = (i_size_read(inode) + blocksize - 1) >> | |
1183 | blkbits; | |
1184 | if (last_block > last_block_in_file) | |
1185 | last_block = last_block_in_file; | |
1186 | ||
1187 | /* | |
1188 | * Map blocks using the previous result first. | |
1189 | */ | |
1190 | if ((map.m_flags & F2FS_MAP_MAPPED) && | |
1191 | block_in_file > map.m_lblk && | |
1192 | block_in_file < (map.m_lblk + map.m_len)) | |
1193 | goto got_it; | |
1194 | ||
1195 | /* | |
1196 | * Then do more f2fs_map_blocks() calls until we are | |
1197 | * done with this page. | |
1198 | */ | |
1199 | map.m_flags = 0; | |
1200 | ||
1201 | if (block_in_file < last_block) { | |
1202 | map.m_lblk = block_in_file; | |
1203 | map.m_len = last_block - block_in_file; | |
1204 | ||
46c9e141 | 1205 | if (f2fs_map_blocks(inode, &map, 0, |
da85985c | 1206 | F2FS_GET_BLOCK_READ)) |
f1e88660 JK |
1207 | goto set_error_page; |
1208 | } | |
1209 | got_it: | |
1210 | if ((map.m_flags & F2FS_MAP_MAPPED)) { | |
1211 | block_nr = map.m_pblk + block_in_file - map.m_lblk; | |
1212 | SetPageMappedToDisk(page); | |
1213 | ||
1214 | if (!PageUptodate(page) && !cleancache_get_page(page)) { | |
1215 | SetPageUptodate(page); | |
1216 | goto confused; | |
1217 | } | |
1218 | } else { | |
09cbfeaf | 1219 | zero_user_segment(page, 0, PAGE_SIZE); |
237c0790 JK |
1220 | if (!PageUptodate(page)) |
1221 | SetPageUptodate(page); | |
f1e88660 JK |
1222 | unlock_page(page); |
1223 | goto next_page; | |
1224 | } | |
1225 | ||
1226 | /* | |
1227 | * This page will go to BIO. Do we need to send this | |
1228 | * BIO off first? | |
1229 | */ | |
3c62be17 JK |
1230 | if (bio && (last_block_in_bio != block_nr - 1 || |
1231 | !__same_bdev(F2FS_I_SB(inode), block_nr, bio))) { | |
f1e88660 | 1232 | submit_and_realloc: |
4fc29c1a | 1233 | __submit_bio(F2FS_I_SB(inode), bio, DATA); |
f1e88660 JK |
1234 | bio = NULL; |
1235 | } | |
1236 | if (bio == NULL) { | |
78682f79 | 1237 | bio = f2fs_grab_bio(inode, block_nr, nr_pages); |
1d353eb7 JK |
1238 | if (IS_ERR(bio)) { |
1239 | bio = NULL; | |
f1e88660 | 1240 | goto set_error_page; |
4375a336 | 1241 | } |
04d328de | 1242 | bio_set_op_attrs(bio, REQ_OP_READ, 0); |
f1e88660 JK |
1243 | } |
1244 | ||
1245 | if (bio_add_page(bio, page, blocksize, 0) < blocksize) | |
1246 | goto submit_and_realloc; | |
1247 | ||
1248 | last_block_in_bio = block_nr; | |
1249 | goto next_page; | |
1250 | set_error_page: | |
1251 | SetPageError(page); | |
09cbfeaf | 1252 | zero_user_segment(page, 0, PAGE_SIZE); |
f1e88660 JK |
1253 | unlock_page(page); |
1254 | goto next_page; | |
1255 | confused: | |
1256 | if (bio) { | |
4fc29c1a | 1257 | __submit_bio(F2FS_I_SB(inode), bio, DATA); |
f1e88660 JK |
1258 | bio = NULL; |
1259 | } | |
1260 | unlock_page(page); | |
1261 | next_page: | |
1262 | if (pages) | |
09cbfeaf | 1263 | put_page(page); |
f1e88660 JK |
1264 | } |
1265 | BUG_ON(pages && !list_empty(pages)); | |
1266 | if (bio) | |
4fc29c1a | 1267 | __submit_bio(F2FS_I_SB(inode), bio, DATA); |
f1e88660 JK |
1268 | return 0; |
1269 | } | |
1270 | ||
eb47b800 JK |
1271 | static int f2fs_read_data_page(struct file *file, struct page *page) |
1272 | { | |
9ffe0fb5 | 1273 | struct inode *inode = page->mapping->host; |
b3d208f9 | 1274 | int ret = -EAGAIN; |
9ffe0fb5 | 1275 | |
c20e89cd CY |
1276 | trace_f2fs_readpage(page, DATA); |
1277 | ||
e1c42045 | 1278 | /* If the file has inline data, try to read it directly */ |
9ffe0fb5 HL |
1279 | if (f2fs_has_inline_data(inode)) |
1280 | ret = f2fs_read_inline_data(inode, page); | |
b3d208f9 | 1281 | if (ret == -EAGAIN) |
f1e88660 | 1282 | ret = f2fs_mpage_readpages(page->mapping, NULL, page, 1); |
9ffe0fb5 | 1283 | return ret; |
eb47b800 JK |
1284 | } |
1285 | ||
1286 | static int f2fs_read_data_pages(struct file *file, | |
1287 | struct address_space *mapping, | |
1288 | struct list_head *pages, unsigned nr_pages) | |
1289 | { | |
9ffe0fb5 | 1290 | struct inode *inode = file->f_mapping->host; |
939afa94 | 1291 | struct page *page = list_last_entry(pages, struct page, lru); |
b8c29400 CY |
1292 | |
1293 | trace_f2fs_readpages(inode, page, nr_pages); | |
9ffe0fb5 HL |
1294 | |
1295 | /* If the file has inline data, skip readpages */ | |
1296 | if (f2fs_has_inline_data(inode)) | |
1297 | return 0; | |
1298 | ||
f1e88660 | 1299 | return f2fs_mpage_readpages(mapping, pages, NULL, nr_pages); |
eb47b800 JK |
1300 | } |
1301 | ||
05ca3632 | 1302 | int do_write_data_page(struct f2fs_io_info *fio) |
eb47b800 | 1303 | { |
05ca3632 | 1304 | struct page *page = fio->page; |
eb47b800 | 1305 | struct inode *inode = page->mapping->host; |
eb47b800 JK |
1306 | struct dnode_of_data dn; |
1307 | int err = 0; | |
1308 | ||
1309 | set_new_dnode(&dn, inode, NULL, NULL, 0); | |
266e97a8 | 1310 | err = get_dnode_of_data(&dn, page->index, LOOKUP_NODE); |
eb47b800 JK |
1311 | if (err) |
1312 | return err; | |
1313 | ||
28bc106b | 1314 | fio->old_blkaddr = dn.data_blkaddr; |
eb47b800 JK |
1315 | |
1316 | /* This page is already truncated */ | |
7a9d7548 | 1317 | if (fio->old_blkaddr == NULL_ADDR) { |
2bca1e23 | 1318 | ClearPageUptodate(page); |
eb47b800 | 1319 | goto out_writepage; |
2bca1e23 | 1320 | } |
eb47b800 | 1321 | |
4375a336 | 1322 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) { |
b32e4482 | 1323 | gfp_t gfp_flags = GFP_NOFS; |
08b39fbd CY |
1324 | |
1325 | /* wait for GCed encrypted page writeback */ | |
1326 | f2fs_wait_on_encrypted_page_writeback(F2FS_I_SB(inode), | |
7a9d7548 | 1327 | fio->old_blkaddr); |
b32e4482 JK |
1328 | retry_encrypt: |
1329 | fio->encrypted_page = fscrypt_encrypt_page(inode, fio->page, | |
9c4bb8a3 DG |
1330 | PAGE_SIZE, 0, |
1331 | fio->page->index, | |
1332 | gfp_flags); | |
4375a336 JK |
1333 | if (IS_ERR(fio->encrypted_page)) { |
1334 | err = PTR_ERR(fio->encrypted_page); | |
b32e4482 JK |
1335 | if (err == -ENOMEM) { |
1336 | /* flush pending ios and wait for a while */ | |
1337 | f2fs_flush_merged_bios(F2FS_I_SB(inode)); | |
1338 | congestion_wait(BLK_RW_ASYNC, HZ/50); | |
1339 | gfp_flags |= __GFP_NOFAIL; | |
1340 | err = 0; | |
1341 | goto retry_encrypt; | |
1342 | } | |
4375a336 JK |
1343 | goto out_writepage; |
1344 | } | |
1345 | } | |
1346 | ||
eb47b800 JK |
1347 | set_page_writeback(page); |
1348 | ||
1349 | /* | |
1350 | * If current allocation needs SSR, | |
1351 | * it had better in-place writes for updated data. | |
1352 | */ | |
7a9d7548 | 1353 | if (unlikely(fio->old_blkaddr != NEW_ADDR && |
b25958b6 | 1354 | !is_cold_data(page) && |
2da3e027 | 1355 | !IS_ATOMIC_WRITTEN_PAGE(page) && |
b25958b6 | 1356 | need_inplace_update(inode))) { |
05ca3632 | 1357 | rewrite_data_page(fio); |
91942321 | 1358 | set_inode_flag(inode, FI_UPDATE_WRITE); |
8ce67cb0 | 1359 | trace_f2fs_do_write_data_page(page, IPU); |
eb47b800 | 1360 | } else { |
05ca3632 | 1361 | write_data_page(&dn, fio); |
8ce67cb0 | 1362 | trace_f2fs_do_write_data_page(page, OPU); |
91942321 | 1363 | set_inode_flag(inode, FI_APPEND_WRITE); |
3c6c2beb | 1364 | if (page->index == 0) |
91942321 | 1365 | set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN); |
eb47b800 JK |
1366 | } |
1367 | out_writepage: | |
1368 | f2fs_put_dnode(&dn); | |
1369 | return err; | |
1370 | } | |
1371 | ||
d68f735b JK |
1372 | static int __write_data_page(struct page *page, bool *submitted, |
1373 | struct writeback_control *wbc) | |
eb47b800 JK |
1374 | { |
1375 | struct inode *inode = page->mapping->host; | |
4081363f | 1376 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
eb47b800 JK |
1377 | loff_t i_size = i_size_read(inode); |
1378 | const pgoff_t end_index = ((unsigned long long) i_size) | |
09cbfeaf | 1379 | >> PAGE_SHIFT; |
26de9b11 | 1380 | loff_t psize = (page->index + 1) << PAGE_SHIFT; |
9ffe0fb5 | 1381 | unsigned offset = 0; |
39936837 | 1382 | bool need_balance_fs = false; |
eb47b800 | 1383 | int err = 0; |
458e6197 | 1384 | struct f2fs_io_info fio = { |
05ca3632 | 1385 | .sbi = sbi, |
458e6197 | 1386 | .type = DATA, |
04d328de | 1387 | .op = REQ_OP_WRITE, |
7637241e | 1388 | .op_flags = wbc_to_write_flags(wbc), |
05ca3632 | 1389 | .page = page, |
4375a336 | 1390 | .encrypted_page = NULL, |
d68f735b | 1391 | .submitted = false, |
458e6197 | 1392 | }; |
eb47b800 | 1393 | |
ecda0de3 CY |
1394 | trace_f2fs_writepage(page, DATA); |
1395 | ||
eb47b800 | 1396 | if (page->index < end_index) |
39936837 | 1397 | goto write; |
eb47b800 JK |
1398 | |
1399 | /* | |
1400 | * If the offset is out-of-range of file size, | |
1401 | * this page does not have to be written to disk. | |
1402 | */ | |
09cbfeaf | 1403 | offset = i_size & (PAGE_SIZE - 1); |
76f60268 | 1404 | if ((page->index >= end_index + 1) || !offset) |
39936837 | 1405 | goto out; |
eb47b800 | 1406 | |
09cbfeaf | 1407 | zero_user_segment(page, offset, PAGE_SIZE); |
39936837 | 1408 | write: |
caf0047e | 1409 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
eb47b800 | 1410 | goto redirty_out; |
1e84371f JK |
1411 | if (f2fs_is_drop_cache(inode)) |
1412 | goto out; | |
e6e5f561 JK |
1413 | /* we should not write 0'th page having journal header */ |
1414 | if (f2fs_is_volatile_file(inode) && (!page->index || | |
1415 | (!wbc->for_reclaim && | |
1416 | available_free_memory(sbi, BASE_CHECK)))) | |
1e84371f | 1417 | goto redirty_out; |
eb47b800 | 1418 | |
cf779cab JK |
1419 | /* we should bypass data pages to proceed the kworkder jobs */ |
1420 | if (unlikely(f2fs_cp_error(sbi))) { | |
7f319975 | 1421 | mapping_set_error(page->mapping, -EIO); |
a7ffdbe2 | 1422 | goto out; |
cf779cab JK |
1423 | } |
1424 | ||
39936837 | 1425 | /* Dentry blocks are controlled by checkpoint */ |
eb47b800 | 1426 | if (S_ISDIR(inode->i_mode)) { |
05ca3632 | 1427 | err = do_write_data_page(&fio); |
8618b881 JK |
1428 | goto done; |
1429 | } | |
9ffe0fb5 | 1430 | |
8618b881 | 1431 | if (!wbc->for_reclaim) |
39936837 | 1432 | need_balance_fs = true; |
7f3037a5 | 1433 | else if (has_not_enough_free_secs(sbi, 0, 0)) |
39936837 | 1434 | goto redirty_out; |
eb47b800 | 1435 | |
b3d208f9 | 1436 | err = -EAGAIN; |
dd7b2333 | 1437 | if (f2fs_has_inline_data(inode)) { |
b3d208f9 | 1438 | err = f2fs_write_inline_data(inode, page); |
dd7b2333 YH |
1439 | if (!err) |
1440 | goto out; | |
1441 | } | |
1442 | f2fs_lock_op(sbi); | |
b3d208f9 | 1443 | if (err == -EAGAIN) |
05ca3632 | 1444 | err = do_write_data_page(&fio); |
26de9b11 JK |
1445 | if (F2FS_I(inode)->last_disk_size < psize) |
1446 | F2FS_I(inode)->last_disk_size = psize; | |
8618b881 JK |
1447 | f2fs_unlock_op(sbi); |
1448 | done: | |
1449 | if (err && err != -ENOENT) | |
1450 | goto redirty_out; | |
eb47b800 | 1451 | |
39936837 | 1452 | out: |
a7ffdbe2 | 1453 | inode_dec_dirty_pages(inode); |
2bca1e23 JK |
1454 | if (err) |
1455 | ClearPageUptodate(page); | |
0c3a5797 CY |
1456 | |
1457 | if (wbc->for_reclaim) { | |
942fd319 JK |
1458 | f2fs_submit_merged_bio_cond(sbi, inode, 0, page->index, |
1459 | DATA, WRITE); | |
0c3a5797 | 1460 | remove_dirty_inode(inode); |
d68f735b | 1461 | submitted = NULL; |
0c3a5797 CY |
1462 | } |
1463 | ||
eb47b800 | 1464 | unlock_page(page); |
2c4db1a6 | 1465 | f2fs_balance_fs(sbi, need_balance_fs); |
0c3a5797 | 1466 | |
d68f735b | 1467 | if (unlikely(f2fs_cp_error(sbi))) { |
2aea39ec | 1468 | f2fs_submit_merged_bio(sbi, DATA, WRITE); |
d68f735b JK |
1469 | submitted = NULL; |
1470 | } | |
1471 | ||
1472 | if (submitted) | |
1473 | *submitted = fio.submitted; | |
0c3a5797 | 1474 | |
eb47b800 JK |
1475 | return 0; |
1476 | ||
eb47b800 | 1477 | redirty_out: |
76f60268 | 1478 | redirty_page_for_writepage(wbc, page); |
0002b61b CY |
1479 | if (!err) |
1480 | return AOP_WRITEPAGE_ACTIVATE; | |
b230e6ca JK |
1481 | unlock_page(page); |
1482 | return err; | |
fa9150a8 NJ |
1483 | } |
1484 | ||
f566bae8 JK |
1485 | static int f2fs_write_data_page(struct page *page, |
1486 | struct writeback_control *wbc) | |
1487 | { | |
d68f735b | 1488 | return __write_data_page(page, NULL, wbc); |
f566bae8 JK |
1489 | } |
1490 | ||
8f46dcae CY |
1491 | /* |
1492 | * This function was copied from write_cche_pages from mm/page-writeback.c. | |
1493 | * The major change is making write step of cold data page separately from | |
1494 | * warm/hot data page. | |
1495 | */ | |
1496 | static int f2fs_write_cache_pages(struct address_space *mapping, | |
b230e6ca | 1497 | struct writeback_control *wbc) |
8f46dcae CY |
1498 | { |
1499 | int ret = 0; | |
1500 | int done = 0; | |
1501 | struct pagevec pvec; | |
1502 | int nr_pages; | |
1503 | pgoff_t uninitialized_var(writeback_index); | |
1504 | pgoff_t index; | |
1505 | pgoff_t end; /* Inclusive */ | |
1506 | pgoff_t done_index; | |
942fd319 | 1507 | pgoff_t last_idx = ULONG_MAX; |
8f46dcae CY |
1508 | int cycled; |
1509 | int range_whole = 0; | |
1510 | int tag; | |
8f46dcae CY |
1511 | |
1512 | pagevec_init(&pvec, 0); | |
46ae957f | 1513 | |
8f46dcae CY |
1514 | if (wbc->range_cyclic) { |
1515 | writeback_index = mapping->writeback_index; /* prev offset */ | |
1516 | index = writeback_index; | |
1517 | if (index == 0) | |
1518 | cycled = 1; | |
1519 | else | |
1520 | cycled = 0; | |
1521 | end = -1; | |
1522 | } else { | |
09cbfeaf KS |
1523 | index = wbc->range_start >> PAGE_SHIFT; |
1524 | end = wbc->range_end >> PAGE_SHIFT; | |
8f46dcae CY |
1525 | if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX) |
1526 | range_whole = 1; | |
1527 | cycled = 1; /* ignore range_cyclic tests */ | |
1528 | } | |
1529 | if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages) | |
1530 | tag = PAGECACHE_TAG_TOWRITE; | |
1531 | else | |
1532 | tag = PAGECACHE_TAG_DIRTY; | |
1533 | retry: | |
1534 | if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages) | |
1535 | tag_pages_for_writeback(mapping, index, end); | |
1536 | done_index = index; | |
1537 | while (!done && (index <= end)) { | |
1538 | int i; | |
1539 | ||
1540 | nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag, | |
1541 | min(end - index, (pgoff_t)PAGEVEC_SIZE - 1) + 1); | |
1542 | if (nr_pages == 0) | |
1543 | break; | |
1544 | ||
1545 | for (i = 0; i < nr_pages; i++) { | |
1546 | struct page *page = pvec.pages[i]; | |
d68f735b | 1547 | bool submitted = false; |
8f46dcae CY |
1548 | |
1549 | if (page->index > end) { | |
1550 | done = 1; | |
1551 | break; | |
1552 | } | |
1553 | ||
1554 | done_index = page->index; | |
1555 | ||
1556 | lock_page(page); | |
1557 | ||
1558 | if (unlikely(page->mapping != mapping)) { | |
1559 | continue_unlock: | |
1560 | unlock_page(page); | |
1561 | continue; | |
1562 | } | |
1563 | ||
1564 | if (!PageDirty(page)) { | |
1565 | /* someone wrote it for us */ | |
1566 | goto continue_unlock; | |
1567 | } | |
1568 | ||
8f46dcae CY |
1569 | if (PageWriteback(page)) { |
1570 | if (wbc->sync_mode != WB_SYNC_NONE) | |
fec1d657 JK |
1571 | f2fs_wait_on_page_writeback(page, |
1572 | DATA, true); | |
8f46dcae CY |
1573 | else |
1574 | goto continue_unlock; | |
1575 | } | |
1576 | ||
1577 | BUG_ON(PageWriteback(page)); | |
1578 | if (!clear_page_dirty_for_io(page)) | |
1579 | goto continue_unlock; | |
1580 | ||
d68f735b | 1581 | ret = __write_data_page(page, &submitted, wbc); |
8f46dcae | 1582 | if (unlikely(ret)) { |
0002b61b CY |
1583 | /* |
1584 | * keep nr_to_write, since vfs uses this to | |
1585 | * get # of written pages. | |
1586 | */ | |
1587 | if (ret == AOP_WRITEPAGE_ACTIVATE) { | |
1588 | unlock_page(page); | |
1589 | ret = 0; | |
1590 | continue; | |
1591 | } | |
b230e6ca JK |
1592 | done_index = page->index + 1; |
1593 | done = 1; | |
1594 | break; | |
d68f735b | 1595 | } else if (submitted) { |
942fd319 | 1596 | last_idx = page->index; |
8f46dcae CY |
1597 | } |
1598 | ||
1599 | if (--wbc->nr_to_write <= 0 && | |
1600 | wbc->sync_mode == WB_SYNC_NONE) { | |
1601 | done = 1; | |
1602 | break; | |
1603 | } | |
1604 | } | |
1605 | pagevec_release(&pvec); | |
1606 | cond_resched(); | |
1607 | } | |
1608 | ||
8f46dcae CY |
1609 | if (!cycled && !done) { |
1610 | cycled = 1; | |
1611 | index = 0; | |
1612 | end = writeback_index - 1; | |
1613 | goto retry; | |
1614 | } | |
1615 | if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) | |
1616 | mapping->writeback_index = done_index; | |
1617 | ||
942fd319 | 1618 | if (last_idx != ULONG_MAX) |
6ca56ca4 | 1619 | f2fs_submit_merged_bio_cond(F2FS_M_SB(mapping), mapping->host, |
942fd319 | 1620 | 0, last_idx, DATA, WRITE); |
6ca56ca4 | 1621 | |
8f46dcae CY |
1622 | return ret; |
1623 | } | |
1624 | ||
25ca923b | 1625 | static int f2fs_write_data_pages(struct address_space *mapping, |
eb47b800 JK |
1626 | struct writeback_control *wbc) |
1627 | { | |
1628 | struct inode *inode = mapping->host; | |
4081363f | 1629 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
9dfa1baf | 1630 | struct blk_plug plug; |
eb47b800 | 1631 | int ret; |
eb47b800 | 1632 | |
cfb185a1 | 1633 | /* deal with chardevs and other special file */ |
1634 | if (!mapping->a_ops->writepage) | |
1635 | return 0; | |
1636 | ||
6a290544 CY |
1637 | /* skip writing if there is no dirty page in this inode */ |
1638 | if (!get_dirty_pages(inode) && wbc->sync_mode == WB_SYNC_NONE) | |
1639 | return 0; | |
1640 | ||
a1257023 JK |
1641 | if (S_ISDIR(inode->i_mode) && wbc->sync_mode == WB_SYNC_NONE && |
1642 | get_dirty_pages(inode) < nr_pages_to_skip(sbi, DATA) && | |
1643 | available_free_memory(sbi, DIRTY_DENTS)) | |
1644 | goto skip_write; | |
1645 | ||
d323d005 | 1646 | /* skip writing during file defragment */ |
91942321 | 1647 | if (is_inode_flag_set(inode, FI_DO_DEFRAG)) |
d323d005 CY |
1648 | goto skip_write; |
1649 | ||
d5669f7b JK |
1650 | /* during POR, we don't need to trigger writepage at all. */ |
1651 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) | |
1652 | goto skip_write; | |
1653 | ||
d31c7c3f YH |
1654 | trace_f2fs_writepages(mapping->host, wbc, DATA); |
1655 | ||
9dfa1baf | 1656 | blk_start_plug(&plug); |
b230e6ca | 1657 | ret = f2fs_write_cache_pages(mapping, wbc); |
9dfa1baf | 1658 | blk_finish_plug(&plug); |
28ea6162 JK |
1659 | /* |
1660 | * if some pages were truncated, we cannot guarantee its mapping->host | |
1661 | * to detect pending bios. | |
1662 | */ | |
458e6197 | 1663 | |
c227f912 | 1664 | remove_dirty_inode(inode); |
eb47b800 | 1665 | return ret; |
d3baf95d JK |
1666 | |
1667 | skip_write: | |
a7ffdbe2 | 1668 | wbc->pages_skipped += get_dirty_pages(inode); |
d31c7c3f | 1669 | trace_f2fs_writepages(mapping->host, wbc, DATA); |
d3baf95d | 1670 | return 0; |
eb47b800 JK |
1671 | } |
1672 | ||
3aab8f82 CY |
1673 | static void f2fs_write_failed(struct address_space *mapping, loff_t to) |
1674 | { | |
1675 | struct inode *inode = mapping->host; | |
819d9153 | 1676 | loff_t i_size = i_size_read(inode); |
3aab8f82 | 1677 | |
819d9153 JK |
1678 | if (to > i_size) { |
1679 | truncate_pagecache(inode, i_size); | |
1680 | truncate_blocks(inode, i_size, true); | |
3aab8f82 CY |
1681 | } |
1682 | } | |
1683 | ||
2aadac08 JK |
1684 | static int prepare_write_begin(struct f2fs_sb_info *sbi, |
1685 | struct page *page, loff_t pos, unsigned len, | |
1686 | block_t *blk_addr, bool *node_changed) | |
1687 | { | |
1688 | struct inode *inode = page->mapping->host; | |
1689 | pgoff_t index = page->index; | |
1690 | struct dnode_of_data dn; | |
1691 | struct page *ipage; | |
b4d07a3e | 1692 | bool locked = false; |
e15882b6 | 1693 | struct extent_info ei = {0,0,0}; |
2aadac08 JK |
1694 | int err = 0; |
1695 | ||
24b84912 JK |
1696 | /* |
1697 | * we already allocated all the blocks, so we don't need to get | |
1698 | * the block addresses when there is no need to fill the page. | |
1699 | */ | |
dc91de78 JK |
1700 | if (!f2fs_has_inline_data(inode) && len == PAGE_SIZE && |
1701 | !is_inode_flag_set(inode, FI_NO_PREALLOC)) | |
24b84912 JK |
1702 | return 0; |
1703 | ||
b4d07a3e | 1704 | if (f2fs_has_inline_data(inode) || |
09cbfeaf | 1705 | (pos & PAGE_MASK) >= i_size_read(inode)) { |
59c9081b | 1706 | __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true); |
b4d07a3e JK |
1707 | locked = true; |
1708 | } | |
1709 | restart: | |
2aadac08 JK |
1710 | /* check inline_data */ |
1711 | ipage = get_node_page(sbi, inode->i_ino); | |
1712 | if (IS_ERR(ipage)) { | |
1713 | err = PTR_ERR(ipage); | |
1714 | goto unlock_out; | |
1715 | } | |
1716 | ||
1717 | set_new_dnode(&dn, inode, ipage, ipage, 0); | |
1718 | ||
1719 | if (f2fs_has_inline_data(inode)) { | |
1720 | if (pos + len <= MAX_INLINE_DATA) { | |
1721 | read_inline_data(page, ipage); | |
91942321 | 1722 | set_inode_flag(inode, FI_DATA_EXIST); |
ab47036d CY |
1723 | if (inode->i_nlink) |
1724 | set_inline_node(ipage); | |
2aadac08 JK |
1725 | } else { |
1726 | err = f2fs_convert_inline_page(&dn, page); | |
1727 | if (err) | |
b4d07a3e JK |
1728 | goto out; |
1729 | if (dn.data_blkaddr == NULL_ADDR) | |
1730 | err = f2fs_get_block(&dn, index); | |
1731 | } | |
1732 | } else if (locked) { | |
1733 | err = f2fs_get_block(&dn, index); | |
1734 | } else { | |
1735 | if (f2fs_lookup_extent_cache(inode, index, &ei)) { | |
1736 | dn.data_blkaddr = ei.blk + index - ei.fofs; | |
1737 | } else { | |
b4d07a3e JK |
1738 | /* hole case */ |
1739 | err = get_dnode_of_data(&dn, index, LOOKUP_NODE); | |
4da7bf5a | 1740 | if (err || dn.data_blkaddr == NULL_ADDR) { |
b4d07a3e | 1741 | f2fs_put_dnode(&dn); |
59c9081b YH |
1742 | __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, |
1743 | true); | |
b4d07a3e JK |
1744 | locked = true; |
1745 | goto restart; | |
1746 | } | |
2aadac08 JK |
1747 | } |
1748 | } | |
b4d07a3e | 1749 | |
2aadac08 JK |
1750 | /* convert_inline_page can make node_changed */ |
1751 | *blk_addr = dn.data_blkaddr; | |
1752 | *node_changed = dn.node_changed; | |
b4d07a3e | 1753 | out: |
2aadac08 JK |
1754 | f2fs_put_dnode(&dn); |
1755 | unlock_out: | |
b4d07a3e | 1756 | if (locked) |
59c9081b | 1757 | __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false); |
2aadac08 JK |
1758 | return err; |
1759 | } | |
1760 | ||
eb47b800 JK |
1761 | static int f2fs_write_begin(struct file *file, struct address_space *mapping, |
1762 | loff_t pos, unsigned len, unsigned flags, | |
1763 | struct page **pagep, void **fsdata) | |
1764 | { | |
1765 | struct inode *inode = mapping->host; | |
4081363f | 1766 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
86531d6b | 1767 | struct page *page = NULL; |
09cbfeaf | 1768 | pgoff_t index = ((unsigned long long) pos) >> PAGE_SHIFT; |
2aadac08 JK |
1769 | bool need_balance = false; |
1770 | block_t blkaddr = NULL_ADDR; | |
eb47b800 JK |
1771 | int err = 0; |
1772 | ||
62aed044 CY |
1773 | trace_f2fs_write_begin(inode, pos, len, flags); |
1774 | ||
5f727395 JK |
1775 | /* |
1776 | * We should check this at this moment to avoid deadlock on inode page | |
1777 | * and #0 page. The locking rule for inline_data conversion should be: | |
1778 | * lock_page(page #0) -> lock_page(inode_page) | |
1779 | */ | |
1780 | if (index != 0) { | |
1781 | err = f2fs_convert_inline_inode(inode); | |
1782 | if (err) | |
1783 | goto fail; | |
1784 | } | |
afcb7ca0 | 1785 | repeat: |
86d54795 JK |
1786 | /* |
1787 | * Do not use grab_cache_page_write_begin() to avoid deadlock due to | |
1788 | * wait_for_stable_page. Will wait that below with our IO control. | |
1789 | */ | |
1790 | page = pagecache_get_page(mapping, index, | |
1791 | FGP_LOCK | FGP_WRITE | FGP_CREAT, GFP_NOFS); | |
3aab8f82 CY |
1792 | if (!page) { |
1793 | err = -ENOMEM; | |
1794 | goto fail; | |
1795 | } | |
d5f66990 | 1796 | |
eb47b800 JK |
1797 | *pagep = page; |
1798 | ||
2aadac08 JK |
1799 | err = prepare_write_begin(sbi, page, pos, len, |
1800 | &blkaddr, &need_balance); | |
9ba69cf9 | 1801 | if (err) |
2aadac08 | 1802 | goto fail; |
9ba69cf9 | 1803 | |
7f3037a5 | 1804 | if (need_balance && has_not_enough_free_secs(sbi, 0, 0)) { |
2a340760 | 1805 | unlock_page(page); |
2c4db1a6 | 1806 | f2fs_balance_fs(sbi, true); |
2a340760 JK |
1807 | lock_page(page); |
1808 | if (page->mapping != mapping) { | |
1809 | /* The page got truncated from under us */ | |
1810 | f2fs_put_page(page, 1); | |
1811 | goto repeat; | |
1812 | } | |
1813 | } | |
1814 | ||
fec1d657 | 1815 | f2fs_wait_on_page_writeback(page, DATA, false); |
b3d208f9 | 1816 | |
08b39fbd CY |
1817 | /* wait for GCed encrypted page writeback */ |
1818 | if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode)) | |
2aadac08 | 1819 | f2fs_wait_on_encrypted_page_writeback(sbi, blkaddr); |
08b39fbd | 1820 | |
649d7df2 JK |
1821 | if (len == PAGE_SIZE || PageUptodate(page)) |
1822 | return 0; | |
eb47b800 | 1823 | |
746e2403 YH |
1824 | if (!(pos & (PAGE_SIZE - 1)) && (pos + len) >= i_size_read(inode)) { |
1825 | zero_user_segment(page, len, PAGE_SIZE); | |
1826 | return 0; | |
1827 | } | |
1828 | ||
2aadac08 | 1829 | if (blkaddr == NEW_ADDR) { |
09cbfeaf | 1830 | zero_user_segment(page, 0, PAGE_SIZE); |
649d7df2 | 1831 | SetPageUptodate(page); |
eb47b800 | 1832 | } else { |
78682f79 | 1833 | struct bio *bio; |
d54c795b | 1834 | |
78682f79 CY |
1835 | bio = f2fs_grab_bio(inode, blkaddr, 1); |
1836 | if (IS_ERR(bio)) { | |
1837 | err = PTR_ERR(bio); | |
3aab8f82 | 1838 | goto fail; |
eb47b800 | 1839 | } |
70fd7614 | 1840 | bio->bi_opf = REQ_OP_READ; |
78682f79 CY |
1841 | if (bio_add_page(bio, page, PAGE_SIZE, 0) < PAGE_SIZE) { |
1842 | bio_put(bio); | |
1843 | err = -EFAULT; | |
1844 | goto fail; | |
1845 | } | |
1846 | ||
4fc29c1a | 1847 | __submit_bio(sbi, bio, DATA); |
d54c795b | 1848 | |
393ff91f | 1849 | lock_page(page); |
6bacf52f | 1850 | if (unlikely(page->mapping != mapping)) { |
afcb7ca0 JK |
1851 | f2fs_put_page(page, 1); |
1852 | goto repeat; | |
eb47b800 | 1853 | } |
1563ac75 CY |
1854 | if (unlikely(!PageUptodate(page))) { |
1855 | err = -EIO; | |
1856 | goto fail; | |
4375a336 | 1857 | } |
eb47b800 | 1858 | } |
eb47b800 | 1859 | return 0; |
9ba69cf9 | 1860 | |
3aab8f82 | 1861 | fail: |
86531d6b | 1862 | f2fs_put_page(page, 1); |
3aab8f82 CY |
1863 | f2fs_write_failed(mapping, pos + len); |
1864 | return err; | |
eb47b800 JK |
1865 | } |
1866 | ||
a1dd3c13 JK |
1867 | static int f2fs_write_end(struct file *file, |
1868 | struct address_space *mapping, | |
1869 | loff_t pos, unsigned len, unsigned copied, | |
1870 | struct page *page, void *fsdata) | |
1871 | { | |
1872 | struct inode *inode = page->mapping->host; | |
1873 | ||
dfb2bf38 CY |
1874 | trace_f2fs_write_end(inode, pos, len, copied); |
1875 | ||
649d7df2 JK |
1876 | /* |
1877 | * This should be come from len == PAGE_SIZE, and we expect copied | |
1878 | * should be PAGE_SIZE. Otherwise, we treat it with zero copied and | |
1879 | * let generic_perform_write() try to copy data again through copied=0. | |
1880 | */ | |
1881 | if (!PageUptodate(page)) { | |
746e2403 | 1882 | if (unlikely(copied != len)) |
649d7df2 JK |
1883 | copied = 0; |
1884 | else | |
1885 | SetPageUptodate(page); | |
1886 | } | |
1887 | if (!copied) | |
1888 | goto unlock_out; | |
1889 | ||
34ba94ba | 1890 | set_page_dirty(page); |
a1dd3c13 | 1891 | |
fc9581c8 JK |
1892 | if (pos + copied > i_size_read(inode)) |
1893 | f2fs_i_size_write(inode, pos + copied); | |
649d7df2 | 1894 | unlock_out: |
3024c9a1 | 1895 | f2fs_put_page(page, 1); |
d0239e1b | 1896 | f2fs_update_time(F2FS_I_SB(inode), REQ_TIME); |
a1dd3c13 JK |
1897 | return copied; |
1898 | } | |
1899 | ||
6f673763 OS |
1900 | static int check_direct_IO(struct inode *inode, struct iov_iter *iter, |
1901 | loff_t offset) | |
944fcfc1 JK |
1902 | { |
1903 | unsigned blocksize_mask = inode->i_sb->s_blocksize - 1; | |
944fcfc1 | 1904 | |
944fcfc1 JK |
1905 | if (offset & blocksize_mask) |
1906 | return -EINVAL; | |
1907 | ||
5b46f25d AV |
1908 | if (iov_iter_alignment(iter) & blocksize_mask) |
1909 | return -EINVAL; | |
1910 | ||
944fcfc1 JK |
1911 | return 0; |
1912 | } | |
1913 | ||
c8b8e32d | 1914 | static ssize_t f2fs_direct_IO(struct kiocb *iocb, struct iov_iter *iter) |
eb47b800 | 1915 | { |
b439b103 | 1916 | struct address_space *mapping = iocb->ki_filp->f_mapping; |
3aab8f82 CY |
1917 | struct inode *inode = mapping->host; |
1918 | size_t count = iov_iter_count(iter); | |
c8b8e32d | 1919 | loff_t offset = iocb->ki_pos; |
82e0a5aa | 1920 | int rw = iov_iter_rw(iter); |
3aab8f82 | 1921 | int err; |
944fcfc1 | 1922 | |
b439b103 | 1923 | err = check_direct_IO(inode, iter, offset); |
b9d777b8 JK |
1924 | if (err) |
1925 | return err; | |
9ffe0fb5 | 1926 | |
c040ff9d | 1927 | if (__force_buffered_io(inode, rw)) |
36abef4e | 1928 | return 0; |
fcc85a4d | 1929 | |
5302fb00 | 1930 | trace_f2fs_direct_IO_enter(inode, offset, count, rw); |
70407fad | 1931 | |
82e0a5aa | 1932 | down_read(&F2FS_I(inode)->dio_rwsem[rw]); |
c8b8e32d | 1933 | err = blockdev_direct_IO(iocb, inode, iter, get_data_block_dio); |
82e0a5aa CY |
1934 | up_read(&F2FS_I(inode)->dio_rwsem[rw]); |
1935 | ||
1936 | if (rw == WRITE) { | |
6bfc4919 | 1937 | if (err > 0) |
91942321 | 1938 | set_inode_flag(inode, FI_UPDATE_WRITE); |
6bfc4919 JK |
1939 | else if (err < 0) |
1940 | f2fs_write_failed(mapping, offset + count); | |
1941 | } | |
70407fad | 1942 | |
5302fb00 | 1943 | trace_f2fs_direct_IO_exit(inode, offset, count, rw, err); |
70407fad | 1944 | |
3aab8f82 | 1945 | return err; |
eb47b800 JK |
1946 | } |
1947 | ||
487261f3 CY |
1948 | void f2fs_invalidate_page(struct page *page, unsigned int offset, |
1949 | unsigned int length) | |
eb47b800 JK |
1950 | { |
1951 | struct inode *inode = page->mapping->host; | |
487261f3 | 1952 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
a7ffdbe2 | 1953 | |
487261f3 | 1954 | if (inode->i_ino >= F2FS_ROOT_INO(sbi) && |
09cbfeaf | 1955 | (offset % PAGE_SIZE || length != PAGE_SIZE)) |
a7ffdbe2 JK |
1956 | return; |
1957 | ||
487261f3 | 1958 | if (PageDirty(page)) { |
933439c8 | 1959 | if (inode->i_ino == F2FS_META_INO(sbi)) { |
487261f3 | 1960 | dec_page_count(sbi, F2FS_DIRTY_META); |
933439c8 | 1961 | } else if (inode->i_ino == F2FS_NODE_INO(sbi)) { |
487261f3 | 1962 | dec_page_count(sbi, F2FS_DIRTY_NODES); |
933439c8 | 1963 | } else { |
487261f3 | 1964 | inode_dec_dirty_pages(inode); |
933439c8 CY |
1965 | remove_dirty_inode(inode); |
1966 | } | |
487261f3 | 1967 | } |
decd36b6 CY |
1968 | |
1969 | /* This is atomic written page, keep Private */ | |
1970 | if (IS_ATOMIC_WRITTEN_PAGE(page)) | |
8c242db9 | 1971 | return drop_inmem_page(inode, page); |
decd36b6 | 1972 | |
23dc974e | 1973 | set_page_private(page, 0); |
eb47b800 JK |
1974 | ClearPagePrivate(page); |
1975 | } | |
1976 | ||
487261f3 | 1977 | int f2fs_release_page(struct page *page, gfp_t wait) |
eb47b800 | 1978 | { |
f68daeeb JK |
1979 | /* If this is dirty page, keep PagePrivate */ |
1980 | if (PageDirty(page)) | |
1981 | return 0; | |
1982 | ||
decd36b6 CY |
1983 | /* This is atomic written page, keep Private */ |
1984 | if (IS_ATOMIC_WRITTEN_PAGE(page)) | |
1985 | return 0; | |
1986 | ||
23dc974e | 1987 | set_page_private(page, 0); |
eb47b800 | 1988 | ClearPagePrivate(page); |
c3850aa1 | 1989 | return 1; |
eb47b800 JK |
1990 | } |
1991 | ||
fe76b796 JK |
1992 | /* |
1993 | * This was copied from __set_page_dirty_buffers which gives higher performance | |
1994 | * in very high speed storages. (e.g., pmem) | |
1995 | */ | |
1996 | void f2fs_set_page_dirty_nobuffers(struct page *page) | |
1997 | { | |
1998 | struct address_space *mapping = page->mapping; | |
1999 | unsigned long flags; | |
2000 | ||
2001 | if (unlikely(!mapping)) | |
2002 | return; | |
2003 | ||
2004 | spin_lock(&mapping->private_lock); | |
2005 | lock_page_memcg(page); | |
2006 | SetPageDirty(page); | |
2007 | spin_unlock(&mapping->private_lock); | |
2008 | ||
2009 | spin_lock_irqsave(&mapping->tree_lock, flags); | |
2010 | WARN_ON_ONCE(!PageUptodate(page)); | |
2011 | account_page_dirtied(page, mapping); | |
2012 | radix_tree_tag_set(&mapping->page_tree, | |
2013 | page_index(page), PAGECACHE_TAG_DIRTY); | |
2014 | spin_unlock_irqrestore(&mapping->tree_lock, flags); | |
2015 | unlock_page_memcg(page); | |
2016 | ||
2017 | __mark_inode_dirty(mapping->host, I_DIRTY_PAGES); | |
2018 | return; | |
2019 | } | |
2020 | ||
eb47b800 JK |
2021 | static int f2fs_set_data_page_dirty(struct page *page) |
2022 | { | |
2023 | struct address_space *mapping = page->mapping; | |
2024 | struct inode *inode = mapping->host; | |
2025 | ||
26c6b887 JK |
2026 | trace_f2fs_set_page_dirty(page, DATA); |
2027 | ||
237c0790 JK |
2028 | if (!PageUptodate(page)) |
2029 | SetPageUptodate(page); | |
34ba94ba | 2030 | |
5fe45743 | 2031 | if (f2fs_is_atomic_file(inode) && !f2fs_is_commit_atomic_write(inode)) { |
decd36b6 CY |
2032 | if (!IS_ATOMIC_WRITTEN_PAGE(page)) { |
2033 | register_inmem_page(inode, page); | |
2034 | return 1; | |
2035 | } | |
2036 | /* | |
2037 | * Previously, this page has been registered, we just | |
2038 | * return here. | |
2039 | */ | |
2040 | return 0; | |
34ba94ba JK |
2041 | } |
2042 | ||
eb47b800 | 2043 | if (!PageDirty(page)) { |
fe76b796 | 2044 | f2fs_set_page_dirty_nobuffers(page); |
a7ffdbe2 | 2045 | update_dirty_page(inode, page); |
eb47b800 JK |
2046 | return 1; |
2047 | } | |
2048 | return 0; | |
2049 | } | |
2050 | ||
c01e54b7 JK |
2051 | static sector_t f2fs_bmap(struct address_space *mapping, sector_t block) |
2052 | { | |
454ae7e5 CY |
2053 | struct inode *inode = mapping->host; |
2054 | ||
1d373a0e JK |
2055 | if (f2fs_has_inline_data(inode)) |
2056 | return 0; | |
2057 | ||
2058 | /* make sure allocating whole blocks */ | |
2059 | if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) | |
2060 | filemap_write_and_wait(mapping); | |
2061 | ||
e2b4e2bc | 2062 | return generic_block_bmap(mapping, block, get_data_block_bmap); |
429511cd CY |
2063 | } |
2064 | ||
5b7a487c WG |
2065 | #ifdef CONFIG_MIGRATION |
2066 | #include <linux/migrate.h> | |
2067 | ||
2068 | int f2fs_migrate_page(struct address_space *mapping, | |
2069 | struct page *newpage, struct page *page, enum migrate_mode mode) | |
2070 | { | |
2071 | int rc, extra_count; | |
2072 | struct f2fs_inode_info *fi = F2FS_I(mapping->host); | |
2073 | bool atomic_written = IS_ATOMIC_WRITTEN_PAGE(page); | |
2074 | ||
2075 | BUG_ON(PageWriteback(page)); | |
2076 | ||
2077 | /* migrating an atomic written page is safe with the inmem_lock hold */ | |
2078 | if (atomic_written && !mutex_trylock(&fi->inmem_lock)) | |
2079 | return -EAGAIN; | |
2080 | ||
2081 | /* | |
2082 | * A reference is expected if PagePrivate set when move mapping, | |
2083 | * however F2FS breaks this for maintaining dirty page counts when | |
2084 | * truncating pages. So here adjusting the 'extra_count' make it work. | |
2085 | */ | |
2086 | extra_count = (atomic_written ? 1 : 0) - page_has_private(page); | |
2087 | rc = migrate_page_move_mapping(mapping, newpage, | |
2088 | page, NULL, mode, extra_count); | |
2089 | if (rc != MIGRATEPAGE_SUCCESS) { | |
2090 | if (atomic_written) | |
2091 | mutex_unlock(&fi->inmem_lock); | |
2092 | return rc; | |
2093 | } | |
2094 | ||
2095 | if (atomic_written) { | |
2096 | struct inmem_pages *cur; | |
2097 | list_for_each_entry(cur, &fi->inmem_pages, list) | |
2098 | if (cur->page == page) { | |
2099 | cur->page = newpage; | |
2100 | break; | |
2101 | } | |
2102 | mutex_unlock(&fi->inmem_lock); | |
2103 | put_page(page); | |
2104 | get_page(newpage); | |
2105 | } | |
2106 | ||
2107 | if (PagePrivate(page)) | |
2108 | SetPagePrivate(newpage); | |
2109 | set_page_private(newpage, page_private(page)); | |
2110 | ||
2111 | migrate_page_copy(newpage, page); | |
2112 | ||
2113 | return MIGRATEPAGE_SUCCESS; | |
2114 | } | |
2115 | #endif | |
2116 | ||
eb47b800 JK |
2117 | const struct address_space_operations f2fs_dblock_aops = { |
2118 | .readpage = f2fs_read_data_page, | |
2119 | .readpages = f2fs_read_data_pages, | |
2120 | .writepage = f2fs_write_data_page, | |
2121 | .writepages = f2fs_write_data_pages, | |
2122 | .write_begin = f2fs_write_begin, | |
a1dd3c13 | 2123 | .write_end = f2fs_write_end, |
eb47b800 | 2124 | .set_page_dirty = f2fs_set_data_page_dirty, |
487261f3 CY |
2125 | .invalidatepage = f2fs_invalidate_page, |
2126 | .releasepage = f2fs_release_page, | |
eb47b800 | 2127 | .direct_IO = f2fs_direct_IO, |
c01e54b7 | 2128 | .bmap = f2fs_bmap, |
5b7a487c WG |
2129 | #ifdef CONFIG_MIGRATION |
2130 | .migratepage = f2fs_migrate_page, | |
2131 | #endif | |
eb47b800 | 2132 | }; |